REPORT: the second burn-down night

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS

.claude/rules/health-checks.md

@@ -21,6 +21,7 @@ fast, and wrong.
 
 This rule used to be duplicated verbatim in felhom-agent/CLAUDE.md with a note explaining that
 felhom.eu/CLAUDE.md "does not load in an agent-only session". That reasoning was correct before
-path-scoped rules existed. The single source is now felhom.eu/CLAUDE.md "Code quality rules"; this
-file is the scoped copy that loads exactly where health checks are written. (2026-08-06)
+path-scoped rules existed. Deliberate scoped copies now live in felhom.eu/.claude/rules/hub.md and
+felhom-controller/.claude/rules/gates.md (hub.md's comment names them); none is the single source. This
+file is the copy that loads exactly where agent health checks are written. (2026-08-06; corrected 2026-10-06)
 -->

.claude/rules/unprompted-work.md

@@ -0,0 +1,69 @@
+---
+unconditional: true
+---
+# Unprompted work — rules for any session without a task file
+
+> Goal sessions, nightly sessions, "work the register" sessions. **A session that starts from
+> `/goal` or a standing brief inherits these rules exactly as it inherits the gates.** They are the
+> part of `PROMPT-TEMPLATE.md` that a task file used to carry and a goal does not. Same wording lives
+> in `felhom.eu`, `felhom-controller`, `felhom-agent` and `app-catalog-felhom.eu` `.claude/rules/`, and in the workspace
+> root's unversioned `.claude/rules/`; change all five or none.
+
+## 1. What you may pick up on your own
+
+- A register row **you or another CC session filed**, with owner CC, at P3 or a bounded P2, that
+  needs **no operator decision**, touches **no customer data by design**, and introduces **no
+  mechanism nobody has measured**. Smallest first.
+- A defect you find while exercising the product, filed as a row **before** you fix it — **unless it is small**:
+  a small finding is fixed in the session and never filed (the size rule, `OPEN-ITEMS.md` „How a row is filed").
+- Hygiene: register compression, stale citations, rows with no owner, documents that contradict
+  live source.
+
+**Not yours, ever, without a task file or an operator word:** money; anything that changes risk to
+customer data; anything that changes a promise the product makes to a customer; anything that
+reverses a documented design decision (`documentation/architecture/` — a design decision is not a
+defect, R-370); anything on DooPlex or ep0; baking or vouching a golden; promoting a
+catalog version; a new external dependency.
+
+## 2. When you may decide instead of ask (operator grant, 2026-09-14)
+
+You may take a decision yourself when **all** of these hold: the architecture folder and the register
+give a clear direction; your choice follows that direction; it is reversible without customer-data
+risk; and you can write it in the `09-update-architecture.md` §3 shape — one answerable sentence, the
+options, what each costs, why this one. **Then record it** as a dated decision in `CONTEXT.md` and
+the owning architecture document, tagged *decided by CC unattended — operator may reverse*, and put
+it **first** in the morning note. A decision you cannot write in that shape is one you do not take.
+
+## 3. The discipline a task file used to carry
+
+1. **Baselines first.** Read each repo's `main` hash and version from live source before touching it.
+2. **Read the architecture document for the area, and name it** in the report, before any claim.
+3. **Red-proof every correctness fix.** A test never seen failing has not been shown to test anything.
+4. **Live-validate on a Tier-0 box** through the endpoints the UI invokes. `demo-hp` is `ssh hp`.
+   Throwaway apps only; the standing apps and `bentopdf` stay.
+5. **Evidence off the machine at the end of each phase**, before any revert (R-320).
+6. **One release per repo per session**, with a CHANGELOG entry (controller: with its `MinAgent`
+   line), REPORT overwritten, floor raised to deliver it. **No golden unless a drill or fresh install
+   needs one** (the waiver, R-468). **No `--no-verify`.**
+7. **An enumerated gap becomes a row in the same session — or, if it is small, is fixed in it** (the size rule).
+   Prose is not a record.
+8. **Hungarian text is searched with ASCII fragments**, with a positive and a negative control.
+9. **Never leave a half-state.** If time runs out, revert to clean and say what was reverted.
+10. **Teardown, three layers, stated** — machine, host, hub — or "provisioned nothing".
+
+## 4. The morning note
+
+One screen, plain language, in this order: **decisions you took** (§2) first; what you exercised;
+what broke and whether you fixed it; rows opened and closed with the register size before and after;
+what needs the operator, each with what happens if they do nothing. No file paths, no function
+names, no row numbers as the subject of a sentence.
+
+## 5. Instruction files
+
+**Instruction files (`CLAUDE.md`, `.claude/rules/*`) are kept true by the session that finds them wrong**
+(operator ruling 2026-10-06, `09` §3 decision 150). A session MAY, without asking: correct a stale fact (a command, a
+count, a version, a path, a description of what a gate does), add a fact it proved, and remove a reference to something
+that no longer exists. Each edit is named in the report (file, line, before, after, why). A session MAY NOT, without the
+operator's word: loosen a safety rule, a fence, a „never", a protected machine, a secret rule, or a review step; or
+remove a rule. When in doubt, it is a rule change, and it goes to the operator. If Claude Code's own permission check
+asks before such an edit, wait for the operator's click; if it refuses, record that and file the exact line.

CHANGELOG.md

@@ -1,3 +1,232 @@
+## Unreleased (2026-10-06 night, later) — after a restart the agent remembers the last backup per tier (R-894); three more SMART counters on the wire (R-330)
+
+Ships with the memory-kill check below as v0.150.0, AFTER the 2026-10-07 night read-back. Nothing delivered tonight.
+
+- **The defect (measured 2026-10-05 on demo-hp):** the agent restarted at 04:57; at 06:25 the off-site storage answered *Can't connect*; the per-tier backup record is in memory only, so the due-check fell back to an EMPTY record and the 7-day tier (last copy 4 days old) read DUE; the controller asked and vzdump failed.
+- New `internal/backup/backup_state.go` `BackupSuccessState`: the newest SUCCESSFUL backup per tier and guest, on disk (`<oob state dir>/backup-success-state.json`, atomic tmp+rename, 0600). Only successes are written; a corrupt file reads as nothing known.
+- `internal/localapi` `handleBackupDue`: when the tier's storage CANNOT be read, the saved copy stands in for the in-memory record. A fresh copy → not due („… (storage unreadable — age from the last success saved on disk)"); a copy older than the cadence → DUE; no copy → the old answer (DUE, age unknown). A storage that answers stays the ground truth: an archive absent there is due even when the file remembers one.
+- Wired in `buildLocalAPIServer` (`LastKnownBackups`); the local API's backup job saves each success.
+- Tests: `TestBackupDue_R894_*` (restart = a new server and a new state from the same file; fresh / old / none / storage answers / failed backup not saved), `TestBackupSuccessState_*`, `TestR894_LastKnownBackupsIsWiredIntoTheDaemon` (AST). Four red-proofs observed (`felhom.eu/documentation/audits/night-burndown-2026-10-06/s4/`).
+- **R-330 (disk health Phase 2, the wire only):** the SMART summary carries three more SATA raw counters — `reported_uncorrect` (187), `command_timeout` (188, carried as the vendor reports it; some pack several counters), `udma_crc_errors` (199). Pointer + omitempty: an attribute the drive does not report is OMITTED (unknown), never 0. No verdict reads them yet. Tests `TestParseSMART_R330_*` (two red-proofs, `felhom.eu/documentation/audits/night-burndown-2026-10-06/r330/`).
+
+## Unreleased (2026-10-06 night) — the Docker step proves the engine reports a memory kill (`09` §3 decision 157, R-528)
+
+To be released as v0.150.0 with its config bundle AFTER the 2026-10-07 night read-back (the night of 2026-10-06 runs v0.149.0 on purpose).
+
+- `configs/felhom-os-apply`: after a docker-layer APPLY (after `health_after`) the wrapper runs `oom_check()`: a throwaway container from the image the running controller uses (`--pull never`, `--network none`, no volume, label `felhom.oomcheck=1`, 64 MB cap) asks for one 200 MB block; „pass" only when `OOMKilled=true` AND the `oom` event; it waits 2 s and reads the events window to the guest's epoch + 1 (measured: a window closed in the same second missed the event); the container is always removed. Reported as `oom_check`; it never changes the step's outcome or health. A wrapper-only mode `oom-check` runs the check alone (no apt, no engine change), by hand as root.
+- `internal/osupdate`: `WrapperReport` and `Report` carry `oom_check` verbatim, on the normal pass and on the kept-copy path (R-868).
+- `configs/test_felhom_os_apply.py`: its `unittest.main()` sat in the middle of the file, so 11 tests (UnsentReport, SaveReportOnDisk, AgentDiesMidPass, CrashLeftTheJournal) never ran — moved to the end; all pass.
+- Tests: the OOMCheck class (pass, OOMKilled=false, no event, unreadable image, removal on an inspect error, not on other layers or in health mode, the events window after the settle wait, mode oom-check alone and its refusals); TestDocker_OOMCheckReachesTheHubUnchanged, TestR868_KeptCopyCarriesTheOOMCheck. 12 red-proofs in `felhom.eu/documentation/audits/readback-2026-10-07/F/`.
+
+## Unreleased (2026-10-06 evening) — the shared rule file (`09` §3 decision 152); no code change
+
+- `.claude/rules/unprompted-work.md` added, byte-identical to the copies in felhom.eu, felhom-controller, app-catalog-felhom.eu and the workspace root (checked with `diff` against the controller's copy and one md5 across all five). Its copies line names five copies.
+
+## Unreleased (2026-10-06 afternoon) — instruction files kept true (`09` §3 decision 150); no code change
+
+- `CLAUDE.md` „Gates — ONE entry point": the runner runs every gate in its `GATES` table (five: three shared, `published`, `release-complete`); `--fast` skips `published` (network). It said two gates and „all of them".
+- `CLAUDE.md`: the decoy gate and its audit are named with their `felhom.eu/` prefix (they do not exist in this repo).
+- `.claude/rules/health-checks.md` (comment): the health-check rule's copies live in felhom.eu `hub.md` and the controller's `gates.md`; it named felhom.eu `CLAUDE.md` „Code quality rules", which holds no such rule.
+
+## v0.149.0 — a weekly disk trim of each customer guest, the crash-boot fact for the controller, the phantom WARN names its runbook (R-444, R-856, R-99; operator rulings `09` §3 139, 143, 140) (2026-10-06)
+
+Released by `scripts/release-agent.sh`: binary sha256 `6bcae9c2eb5d97e8285316583870059835793893299e291891a53a4ce505585f`
+config bundle sha256 `e182c82dcf4a67faa3bcb74dbe4ffa7b06e0b27dc8451cb7574d6339ce91ad66` (tag `v0.149.0` = `f277e61`).
+**The bundle carries the new sudoers rule for the trim (`FELHOM_FSTRIM`) — deliver it with the binary:** signed
+`agent_update`, then signed `agent_config_update`.
+
+- R-856 (`09` §3 decision 143): new local-API route `GET /host/crash-guard` — passes the host crash guard's last-boot record (present, last_boot_at, last_boot_unclean, tripped) from /var/lib/felhom-crash-guard/state.json to the controller, which waits ~15 min with app mails after a crash boot. Read-only, no Proxmox call, guest-token authed; a missing/unreadable/garbled file answers 200 present:false (never an error page). An older agent answers 404, which the controller reads as unknown (normal 90 s grace) — no controller MinAgent raise needed.
+- R-444 (`09` §3 decision 139): weekly guest disk trim. New sudoers alias FELHOM_FSTRIM with ONE exact rule `/usr/sbin/pct ^fstrim [0-9]+$` (rides the signed config bundle; decoys pinned by TestSudoersFstrimRuleIsExact) and capability guest-fstrim (non-critical). New internal/fstrim job: each owned RUNNING guest gets `pct fstrim <vmid>` once a week - due Wednesday from 10:00 host-local, starts only 10:00-20:59 (never the 01:00-06:59 night), holds the one-heavy-op gate so it never runs beside a backup or restore-test (busy -> deferred to the next hourly tick; a box that was off catches up at its next daytime hour); a failed trim WARNs and is retried at most 3 times that week; bytes parsed from `pct fstrim`'s "(N bytes) trimmed" lines; positive log `fstrim: guest N trimmed X GiB in Ys`; last result per guest persisted in <state_dir>/guest-disk-trim.json and reported as the new omitempty host-report stanza `guest_disk_trim`. Opt-out: agent.json "disk_trim": {"disable": true}.
+- R-99 (`09` §3 decision 140): the agent's WARN for a PBS archive below the 1 MiB plausibility floor now ends with the pointer to the sanctioned cleanup (`documentation/runbooks/pbs-phantom-cleanup.md`); detection only — nothing is deleted automatically. Dir-storage archives keep the old text.
+
+## unreleased
+
+- R-426: scripts/test_gate_decoys.py (new) — the published gate judged against a fake Gitea (127.0.0.1, via GITEA_BASE; never the real registry): 11 cases; COVERS published — `felhom-agent/published` leaves the decoy-coverage EXEMPT list.
+- R-426: release-complete gate — 10 decoy cases (scratch clone + scratch bare origin + fake Gitea); COVERS release-complete — `felhom-agent/release-complete` leaves the decoy-coverage EXEMPT list.
+- R-426: the shared reuse-refs/instructions/observations gates get agent-side decoys (9 cases on a scratch clone of this repo); COVERS reuse-refs, instructions, observations — three `felhom-agent/*` entries leave the decoy-coverage EXEMPT list.
+- **Fixed without a row:** `configs/test_felhom_config_bundle.py` read the two ISO first-boot files that installer 1.32.0 now NAMES under KEPT (R-275) as files the installer writes — `go test ./internal/osupdate` was red on DooPlex from 21:25 to 01:55 (felhom.eu `85de3f9b`); they are listed with why. Test only; agent v0.148.0's code is unaffected.
+
+## v0.148.0 — the host report names the running binary's sha; the format answer carries the new filesystem's UUID (burn-down night: R-349, R-25 agent halves) (2026-10-06)
+
+Released by `scripts/release-agent.sh`: binary sha256 `3e68a0870e0e2ce262cb4819294edddeb0a73e8c558a31611a20329a6d9ee283`
+config bundle sha256 `a6fa4f589d184b58c9911303bd087e300be1e75b3647e4302c9594df6989c4de` (tag `v0.148.0` = `861d32a`).
+Delivery order as for v0.147.0: signed `agent_update`, then signed `agent_config_update`.
+
+- **R-349:** the host report carries `agent_sha256`, the sha256 of the running agent binary (read once from
+  `/proc/self/exe`; empty = unknown), so a hand-built binary under the vouched version name becomes visible. The hub
+  comparison is a separate hub change. Test `TestCollect_AgentSHA256IsTheRunningBinary`; red-proved.
+- **R-25 (agent half):** `POST /disks/format` and `GET /disks/format/status` return `fs_uuid`, the new filesystem's UUID
+  read back after mkfs only when the bound durable id still resolves to the formatted device and the superblock is the
+  requested type (empty = not verified); `DeviceProbe` gains `FSUUID` from blkid. Tests `TestFormat_*FSUUID*`; three
+  red-proofs. The controller half (mount that UUID) is a controller change.
+
+## v0.147.0 — the recovery recipe spells the root namespace the way PBS does; a removed drive no longer shows the root disk's size; a rotated-out token stops at once; the dnsmasq check looks at the right package (burn-down round 2: R-124, R-118, R-269, R-317) (2026-10-05)
+
+Released by `scripts/release-agent.sh`: binary sha256 `642c4d196c48671c14ff653118303c5903af1670b7abb550edeaf73e701cd5b8`,
+config bundle sha256 `326527d0993c9a62df2f790c7700ca645cedbf0673dcfb6dc1768d8610b8007d` (tag `v0.147.0` = `f1b9b41`).
+Delivery order as for v0.146.1: signed `agent_update`, then signed `agent_config_update`.
+
+MinAgent impact: none (the controller needs nothing new from this agent). Config bundle content unchanged from v0.146.1.
+
+- **R-124 (operator ruling 2026-10-05: fix it):** the DR recipe's `pbs.namespace` for a box in PBS's ROOT namespace is
+  now `""` — PBS's own spelling — beside `namespace_state: resolved`; it used to be the word `root`, which no namespace
+  is named, so `--ns root` failed in a recovery. `hub.PBSRootNamespace`; `TestR124_RootNamespaceOnTheWireIsPBSSpelling`
+  (red-proof: back to "root" → FAIL). Runbook: `felhom.eu runbooks/ep0-datastore-copy.md` step 2 says how to read it
+  (and to treat a recorded `root` from older agents as empty). The hub stores the recipe raw; its fixture follows.
+- **R-118:** the local API's drive list reads a drive's capacity only while its DEVICE is present — with the device gone
+  the bare mountpoint is a directory on the root filesystem, whose size was reported as the drive's.
+  `TestDisks_UnionPath_AbsentDeviceReportsNoRootCapacity` (red-proof convicts).
+- **R-269:** the token store re-reads its shared file whenever it has grown, BEFORE answering — so a token rotated out by
+  another process stops authorizing on its next use (it used to keep working until an unrelated miss). One `stat` per
+  call. `TestTokenStore_RotatedOutTokenRejectedFirst` (red-proof convicts).
+- **R-317:** the LAN resolver decides whether to install `dnsmasq` by its service UNIT, not by `/usr/sbin/dnsmasq` (which
+  the `dnsmasq-base` package also ships). Same `apt-get install` command; no sudoers change. `TestEnsureDnsmasq_*`
+  (red-proof convicts). Red-proofs: `felhom.eu/documentation/audits/burndown2-2026-10-05/agent-red-proofs.txt`,
+  `r124-red-proof.txt`.
+
+Also in this release (no binary effect; from burn-down round 1):
+
+- **R-291:** `scripts/retention-policy.json` names where its 10 comes from — the R-267 newest-10 prune of generic
+  packages, established 2026-08-10 (R-287) — instead of „observed, no located ruling"; the non-existent
+  `registry-retention.md` reader is dropped. `check-published-versions.py` still reads 10 (checked).
+- **R-348:** `internal/backup/store.go` no longer says backups are „unaffected" by a restart: the reported backup list
+  reads 0 until the next backup runs; only the hub's verdict (7-day look-back) is unaffected.
+
+## v0.146.1 — R-861 review fixes: the signed update flips a root-owned copy; no Wants=/continuations in mount units; the escrow read follows no symlink anywhere (2026-10-05)
+
+Released by `scripts/release-agent.sh`: binary sha256 `badd6c9a2e40c8bfe856d2d1a203443b21b7eb92ecc35d6090ab44518d4d082a`,
+config bundle sha256 `42333e969028867ad8142335e6c1bc4040eec231de0d8d330c2d4b2cf7bc3442`. **Supersedes v0.146.0, which
+was released but never vouched or delivered to any box.** The same order applies: signed `agent_update` first, then the
+signed `agent_config_update`.
+
+**Delivery needs a STEP bundle (R-880, found while delivering).** An installed `felhom-os-apply` checks an incoming
+bundle's paths against its OWN table (R16), so every box on the v0.145.0 bundle REFUSES the v0.146.1 bundle (it adds 4
+paths). `scripts/build-step-bundle.py` builds the transition: the box's current bundle with ONLY `felhom-os-apply`
+replaced (same paths — the old wrapper accepts it), published as bundle version `0.146.1-step1`; then the release's own
+bundle. Order on a box: `agent_update` 0.146.1 → `agent_config_update` 0.146.1-step1 → `agent_config_update` 0.146.1.
+Tests `StepBundle` (the R16 refusal reproduced; the step accepted; exactly one file changed). Tooling only — not in the
+binary or the bundle.
+
+A background security review of the v0.146.0 commit found three holes in the new code; each is fixed and red-proved
+(`felhom.eu/documentation/audits/hub-safety-2026-10-05/partF/red-proof.txt`, S1–S3):
+
+- **S1 — a race in the signed update.** `felhom-os-apply` hashed the agent's staged file and then let the A/B wrapper copy
+  it BY PATH; the agent owns that directory and could swap the file in between. Now the root step reads the file ONCE
+  (`read_staged_once`: O_NOFOLLOW, fstat, owner, size), hashes those bytes, writes them to a root-owned directory
+  (`/var/lib/felhom-os-apply/agent-update/`) and hands ONLY that copy to `felhom-selfupdate-guarded apply`, which now
+  refuses any other directory, a symlink, or a file not owned by root. Tests: `AgentUpdate` (+1),
+  `SelfupdateWrapperConfinement`.
+- **S2 — an allowlist escape in `felhom-priv-apply`.** `[Unit]` accepted `Wants=`/`Requires=`/`Before=` naming any unit, so
+  a mount unit could start e.g. `reboot.target`. `[Unit]` now holds only `Description` and `After=local-fs-pre.target`
+  (what the renderers write), and any line ending in a backslash (a systemd continuation this parser would read
+  differently) is refused. Tests `test_U2_wants_starts_another_unit`, `test_U2_continuation_line`.
+- **S3 — a path traversal in the escrow read.** `O_NOFOLLOW` guards only the last component; a symlinked DIRECTORY in the
+  agent's own state dir still redirected the root read. `readStagedNoFollow` now walks the path from `/` with
+  `openat(O_NOFOLLOW)` per component. Test `TestAttach_RefusesASymlinkedDirectory`.
+
+## v0.146.0 — the agent's root grants narrowed: exact sudo patterns, a root content checker, fixed files from the bundle, the signed update checked as root (R-861) (2026-10-05)
+
+Released by `scripts/release-agent.sh`: binary sha256 `b860af465076041e07f35fed1b12d64ae2b2985d8995f0ce167418d39c2b00d5`,
+config bundle sha256 `161c737e523aa7910cf32ce41b83f989569bee55b8c5938e7211c92aef68548e`. **Order on a box: the signed
+`agent_update` FIRST (the old bundle still grants the old flip), then the signed `agent_config_update`.** Between the
+two (minutes) the new agent's checker calls are refused and retried; nothing is lost. After the bundle, an agent BELOW
+0.146.0 cannot update itself on that box any more (the unsigned flip grant is gone) — deliver both together.
+
+Design: `felhom.eu/documentation/architecture/03-host-agent.md` §3.1 (new). Measured before the change (real sudo
+1.9.16, a throwaway container): the v0.145.0 sudoers let **23 of 29** attack command lines through; v0.146.0 lets
+**0** through and still allows all **64** commands the agent's capability check uses.
+
+- **Exact patterns.** A sudoers `*` in the arguments also matches spaces: `pct set [0-9]* -onboot 1` matched
+  `pct set 100 --dev0 /dev/sda -onboot 1` (a raw host disk for a guest), `mount --bind /mnt/*/felhom-data
+  /mnt/felhom-drives/*` matched a `..` path onto `/etc/sudoers.d`, `nft add element … *` took a chained `; flush
+  ruleset`. Every varying argument list is now a sudo regex (`^…$`): one value per slot, a fixed character set, no
+  `..`, no extra argument. `TestSudoersRefusesTheR861Injections` (29 attacks) + `TestManifestCoveredBySudoers`
+  (regex-aware now).
+- **`felhom-priv-apply`** (new root wrapper, in the bundle). A systemd mount/automount unit, a dnsmasq drop-in, the
+  WireGuard config and the OOB sshd config + felhom-op key reach their root-read places only through it: fixed source,
+  fixed destination, CONTENT checked against what the agent's renderers write (no `[Service]`, `Where=` only
+  `/mnt/<name>` or `/mnt/felhom-drives/<name>` and equal to the unit name, no `bind`/`suid`; a network share must carry
+  `nosuid,nodev`; no `dhcp-script=`; no `PostUp=`; the sshd config only the one template with its Port). Its 30 tests
+  (`configs/test_felhom_priv_apply.py`) + Go contract tests feeding each renderer's real output
+  (`internal/privapplytest`). Pre-flight: every live file on both demo boxes reads OK.
+- **NFS/SMB options gain `nosuid,nodev`** (a set-uid file on a server outside the box never acts on the host).
+- **Fixed files from the bundle.** The guest pre-start hook (`/var/lib/vz/snippets/felhom-guest-hook.sh`, run as root at
+  every guest start) and the shared drive parent script + unit are bundle files now (byte-identical to the agent's
+  constants, pinned). The agent no longer installs them from `/tmp`; it checks them (`guesthook.SnippetReady`,
+  `ensureSharedParentBoot`) and only registers / enables.
+- **The signed update is checked as root.** `felhom-os-apply` mode `agent_update` verifies the operator signature
+  (root-owned signers, this host, the window, the nonce), re-hashes the staged binary against the SIGNED sha, then runs
+  the A/B flip; `felhom-selfupdate-guarded apply` is no longer in the agent's sudoers. 7 tests (`AgentUpdate`).
+- **The root escrow run reads no path from the agent's config.** As root it pins the PVE secret dir and the WireGuard
+  state dir to their defaults, refuses a storage id that is a path, and reads its two staged files without following a
+  symlink (`readStagedNoFollow`) — before, a symlink in the agent's own directory sealed any root file into the blob.
+- **Not narrowed here (named in `03` §3.1):** `FELHOM_CONTROLLERSWAP` stays guest-scoped (a compromised agent can run a
+  chosen controller image in the guest — the household's data, not host root); `FELHOM_ESCROW` still hands the agent R
+  by design (the agent relays the ceremony); the mkfs / pbs-apply / backup-target wrappers keep a coarse argument and
+  their own checks.
+- Red-proofs F1–F9: `felhom.eu/documentation/audits/hub-safety-2026-10-05/partF/red-proof.txt` (F1's first run did NOT
+  convict — the name rule masked it — and the test now uses the pair only the Where rule stops).
+
+## v0.145.0 — the OS update repairs itself after a power cut; a short-session box gets restore-tested; "sent late" (R-876, R-874, R-875) (2026-10-05)
+
+Released by `scripts/release-agent.sh`: binary sha256 `894da35c7b9e1ac78885690b78352b634e6831e7d99b321573c75d878db8886e`,
+config bundle sha256 `78c00adce662d2d966b2ac50ebde46cde1ae225f0107c6a7c02b70ec8ce80c4f`. The wrapper changed: a box
+needs the signed `agent_update` AND the signed `agent_config_update`.
+
+- **R-876.** After a crash during an install, `dpkg --audit` can read clean while dpkg's update journal
+  (`/var/lib/dpkg/updates/`) is not — and apt refuses every install until `dpkg --configure -a` (measured on demo-hp
+  2026-10-05: every later pass failed until a person typed it). The wrapper now reads `--audit` and the journal in ONE
+  `sh -c` call (`DPKG_STATE_SCRIPT`) — a clean pass still costs one call (R-845's speed, pinned) — and repairs when
+  either shows something; as a belt, when apt itself says "dpkg was interrupted", it repairs and retries the install
+  ONCE. `REPAIR` now logs `journal=N`; a journal still not empty after the repair refuses (R13). Tests
+  `CrashLeftTheJournal` (the measured shape, the speed, the belt); 3 red-proofs.
+- **R-874.** The restore-test's first due-check runs 30 minutes after the agent starts (`DefaultFirstEval`), then every
+  interval; a box whose power-on sessions are shorter than the 6 h interval never evaluated. A crash-looping agent
+  restarting faster than 30 minutes still never evaluates (the earned restraint, pinned).
+- **R-875.** A kept report's reason is neutral — "sent late — kept on the box until the hub could take it" — the copy
+  cannot tell a killed agent from an absent hub.
+- Red-proofs: `felhom.eu/documentation/audits/catchup-2026-10-05/part{C,D}/`.
+
+## v0.144.1 — a killed pass really keeps its report: the wrapper survives a dead reader; the agent looks again every 5 minutes (R-868, measured live) (2026-10-05)
+
+Released by `scripts/release-agent.sh`: binary sha256 `6ccd521d47e64999017e8eb5bc613d724543cfdc5ef9b13bcae9e3ea8c53b8f3`,
+config bundle sha256 `e89a9ddfb767e177f8874d56f3dcd3bfd47409d157ff830d362653333bf815e8`. The wrapper changed again:
+a box needs the signed `agent_update` AND the signed `agent_config_update`.
+
+- **Found live on demo-hp 2026-10-05 05:45 UTC with v0.144.0** (the night's A5 shape: kill -9 of the pass and the
+  daemon while apt-get ran): apt finished all 13 packages, but the wrapper's next log line went to a stderr pipe no
+  process read any more → `BrokenPipeError` → the wrapper died before it saved its report copy (journal: `PLAN
+  upgrade=13`, then nothing; no copy; the hub got nothing). v0.144.0's mechanism was right and never reached.
+  `Runner.log` and the final `OSAPPLY-REPORT` line now survive a dead reader (the journal still gets every line).
+  Test `AgentDiesMidPass` drives the REAL `log()` into a pipe that breaks while apt-get runs.
+- **Also found live:** the restarted daemon looked for kept copies ~7 s before the orphaned wrapper wrote one. The
+  daemon now looks at start and every 5 minutes (`Leg.SendUnsentLoop`); `TestR868_ACopyWrittenAfterTheStartIsSentByTheLoop`.
+- Red-proofs: `felhom.eu/documentation/audits/night-fixes-2026-10-05/partD/r868-brokenpipe-red-proof.txt`.
+- A second agent release in one session, against "one release per repo": recorded as `09` decision 108 (operator may
+  reverse) — the alternative was to ship a fix proven not to work.
+
+## v0.144.0 — R8 measures the real download; an OS pass reports even when its agent was killed; the debug pass runs with the hub away (R-865, R-868, R-866) (2026-10-05)
+
+Released by `scripts/release-agent.sh`: binary sha256 `f18093c3466749ec4cd47f83f97a401160a24bad1e704f1183051adad14db928`,
+config bundle `felhom-config-bundle.json` sha256 `6acf42fe46df5223384d767801cb2bf73238ba4dab82debb7811ca2f790591d8`.
+The wrapper `felhom-os-apply` changed, so a box needs BOTH the signed `agent_update` and the signed `agent_config_update`.
+
+- **R-865.** `download_bytes` runs `apt-get --print-uris` WITHOUT `-s`: with `-s` apt prints the simulation and no URI
+  list, so R8 summed 0 B and only its 500 MB floor ever applied. `--print-uris` alone downloads nothing (measured on
+  9202: the archive cache and the versions unchanged). The test fake now answers like real apt (with `-s`: no URIs),
+  and `test_R8_counts_the_real_download` / `test_download_bytes_never_simulates` pin it.
+- **R-868.** The wrapper writes every apply pass's report to `<plan dir>/report-<run>-<layer>-apply.json` before it
+  prints it (root writes into the agent's dir: the dir opened O_NOFOLLOW and checked to be the agent's own, the file
+  created O_EXCL|O_NOFOLLOW, 0600, handed to the agent). The plan now carries `run_id`, `trigger`, `ring`, echoed in
+  the report. The agent deletes the copy once the hub has the report; a copy left on disk (the agent was killed, or
+  the hub was away) is sent at the agent's start and before every pass (`Leg.SendUnsent`), then deleted. A pass lock
+  (flock on `pass.lock`, across the daemon and a selftest) keeps the sender off a pass that is still running.
+- **R-866.** The daemon saves the hub's newest os_update block (`os-update-block.json`); `--selftest=os-update` uses it
+  when the hub cannot be reached and says so in its header (`block=SAVED(<time>; hub unreachable: …)`); with no hub
+  and nothing saved it does not run.
+- Tests: `configs/test_felhom_os_apply.py` (UnsentReport, SaveReportOnDisk — real files, symlink cases), Go
+  `TestR868_*`, `TestR866_*`. Red-proofs: `felhom.eu/documentation/audits/night-fixes-2026-10-05/part{C,D}/`.
+
 ## v0.143.0 — the config bundle: a signed route for a box's root-owned files (R-840, decision 96) (2026-10-04)
 
 Released by `scripts/release-agent.sh`: binary sha256 `41c0d3060013dfda795262147454248149bee0888c0935170fdb85de6e7a35da`,

CLAUDE.md

@@ -52,11 +52,11 @@ This is in the core because breaching it is how this component stops being audit
 
 ## Gates — ONE entry point
 
-**Run `python3 scripts/agent_gates.py` from the repo root after ANY change here.** It runs this
-repo's gates — `reuse_refs_check` and `instructions_gate`, both the **shared** copies in
-`felhom.eu/scripts/`, never copied into this repo (a copy recreates the drift they detect; an absent
-sibling clone FAILS). `--fast` selects the gates touching no network and no container runtime; today
-that is all of them. **A missing gate is a FAILURE, never a skip.**
+**Run `python3 scripts/agent_gates.py` from the repo root after ANY change here.** It runs every
+gate in its `GATES` table (that table is the list); the shared ones — `reuse_refs_check`,
+`instructions_gate`, `observations_gate` — are the copies in `felhom.eu/scripts/`, never copied into
+this repo (a copy recreates the drift they detect; an absent sibling clone FAILS). `--fast` selects the
+gates touching no network and no container runtime, and skips `published` (network), naming it. **A missing gate is a FAILURE, never a skip.**
 
 **The pre-push hook** (`.githooks/pre-push`) runs it with `--fast` and refuses a failing push. It is
 **per-clone** — switch it on once with `git config core.hooksPath .githooks`, and a manual run WARNS
@@ -104,8 +104,8 @@ the mechanism are exempt.
 
 **A gate ships with a decoy test that has been seen to fail (R-421).** A decoy is the LABEL without
 the FACT — a directory with the right name and no bake log, a note whose prose mentions the marker it
-lacks. `scripts/decoy_coverage_gate.py` refuses a new gate that has neither a decoy nor a named
+lacks. `felhom.eu/scripts/decoy_coverage_gate.py` (run by felhom.eu's `repo_gates.py`, for all four repos) refuses a new gate that has neither a decoy nor a named
 exemption carrying its row. The four shapes, the 2026-09-01 sweep that fooled 16 of 29 gates, and the
-decoys withdrawn as illegitimate: `documentation/audits/AUDIT-gate-decoys-2026-09-01.md` and
+decoys withdrawn as illegitimate: `felhom.eu/documentation/audits/AUDIT-gate-decoys-2026-09-01.md` and
 `felhom-controller/.claude/rules/gates.md`. **Scope is a fact too** — prefer `os.walk` over
 `os.listdir`, and a glob over a hand-maintained list.

REPORT.md

@@ -1,19 +1,14 @@
-# REPORT — agent v0.143.0: the config bundle (R-840) — 2026-10-04
+# REPORT — the second burn-down night (2026-10-06 → 07): R-894 and the R-330 wire half, on main, unreleased
 
-**What:** a signed route for a box's root-owned files. `felhom-os-apply` gained mode `bundle` (signed
-`agent_config_update`, verified by the wrapper itself) and `--install-bundle` (the installer's root entry);
-`BUNDLE_FILES` is the one table of 22 paths; `scripts/build-config-bundle.py` builds it reproducibly;
-`release-agent.sh` publishes it beside the binary; the agent reports `system.config_bundle`; `felhom-opsign` signs it.
-Also `build-golden.sh` 3.2.0 (`GOLDEN_GUEST_PKGS`).
+Nothing was released or delivered (brief fence). Both changes ship with the memory-kill check as **v0.150.0**, after the
+2026-10-07 night read-back. Session report: `felhom.eu/REPORT-night-burndown-2026-10-06.md`; night log
+`felhom.eu/documentation/audits/night-burndown-2026-10-06/NIGHT-LOG.md`.
 
-**Released:** v0.143.0, binary `41c0d306…`, bundle `8d7273cf…` (reproducible), vouched in the hub with golden 0.293.0.
-**Delivered:** demo-hp and demo-felhom (signed `agent_update`), then the one-file bootstrap, then the bundle by the
-signed route (0 written of 22; probe 71/71). demo-hp also: a wrong sha refused, a one-line change and its undo, a replay
-rejected. Tester 2: the signed `agent_update` queued (operator ruling 97); its bundle waits for the operator's
-bootstrap (R-862).
-
-**Tests:** `configs/test_felhom_config_bundle.py` 43 (22 of 22 mutants red), Go `internal/osupdate/bundle_test.go`,
-`internal/hub/bundle_record_test.go`; `go vet ./... && go test ./...` rc=0; gates green.
-**Found:** R-861 — the sudoers already lets the agent user reach root (read, not exploited).
-Evidence and the full report: `felhom.eu/documentation/audits/r840-config-bundle-2026-10-04/`,
-`felhom.eu/REPORT-r840-config-bundle-2026-10-04.md`.
+- **R-894** (`74b5eae`): the newest successful backup per tier is kept on disk (`backup-success-state.json` in the OOB
+  state dir) and read by `/backup/due` **only when the tier's storage cannot be read**: fresh → not due, older than the
+  cadence → due, none → due with age unknown (as before). A storage that answers stays the ground truth. Tests
+  `TestBackupDue_R894_*`, `TestBackupSuccessState_*`, `TestR894_LastKnownBackupsIsWiredIntoTheDaemon`; four red-proofs
+  `felhom.eu/documentation/audits/night-burndown-2026-10-06/s4/`. Architecture: `07` §6.1 updated.
+- **R-330, wire only** (`adaf86a`): the host report's SMART summary carries 187 `reported_uncorrect`, 188
+  `command_timeout`, 199 `udma_crc_errors` (pointers, omitted when unknown). No verdict reads them.
+- Suite `go test ./...` rc 0; `scripts/agent_gates.py` (full) all OK; CI green on each push (runs 1451, 1467).

REUSE.md

@@ -14,12 +14,15 @@
 | `Privileged` (CreateGoldenLXC/MountUSBByUUID/SMART/Sensors) | internal/proxmox/privileged.go | methods on `*Privileged` | the 3 fenced root-CLI exceptions ONLY | Do NOT add methods — fence is structural (`routing_test.go` asserts it) |
 | `SudoHostOps.run` | internal/storage/hostops.go | `run(ctx, name, args...) error` | allowlisted exec with stderr-wrapped error | Every arg pre-validated via validate.go before this is called |
 | `Prober.Probe` | internal/capability/probe.go | `Probe(ctx) []Status` | live sudo-policy capability check (`sudo -n -l --`) | Needs a DIRECT runner (never the sudo-prefixing one — double-sudo); never executes probed cmds. v0.86.0: config-gated caps (`Capability.GatedBy` + `Prober.GateActive`) report `inactive`/"disabled by configuration" ONLY when healthy — broken plumbing stays degraded; the pbsdr-* gate answers from `pbsdr.Manager.DRConfigured` (marker-backed across restarts) |
-| `stageTemp` | internal/localapi/intermediary.go | `stageTemp(pattern, content) (path, err)` | random-named temp before a root `install` (audit B1) | Fixed /tmp names are a TOCTOU — sudoers globs expect `/tmp/felhom-*-*.ext` |
-| `BUNDLE_FILES` + `Bundle` (mode `bundle`, `--install-bundle`) | configs/felhom-os-apply | the ONE table of root-owned paths + the installer of them | ANY new root-owned file the installer writes (sudoers line, wrapper, unit) — add it to the table, never a new installer fetch (R-840) | The builder (`scripts/build-config-bundle.py`) and the installer read the same table; `test_every_root_file_the_installer_writes_is_in_the_bundle` fails on a path the bundle lacks. Trust files (`/etc/felhom/os-trust.json`, `operator-signers`) are NEVER bundle paths (R17) |
+| ~~`stageTemp`~~ (REMOVED v0.146.0, R-861) | — | — | — | Nothing the agent writes is `install`ed where root reads it any more: use `felhom-priv-apply` (below) or ship a fixed file in the bundle |
+| `felhom-priv-apply` (v0.146.0, R-861) | configs/felhom-priv-apply | `felhom-priv-apply unit <name> \| dnsmasq <tmp> <name> \| wg \| sshd-config \| sshd-key` | ANY agent-rendered file a root program reads (systemd unit, dnsmasq drop-in, wg-quick conf, OOB sshd) — fixed source + destination, CONTENT checked against the agent's own renderers | A new renderer needs a verb + a contract test (`internal/privapplytest.Check`) feeding its REAL output; never a new `install` sudoers line |
+| `privapplytest.Check` | internal/privapplytest/check.go | `Check(t, verb, name, content) string` | the Go↔root-checker contract: a renderer's real output must read `OK` | Skips without python3; one call per rendered shape + one refused control |
+| `BUNDLE_FILES` + `Bundle` (mode `bundle`, `--install-bundle`; mode `agent_update` v0.146.0, R-861) | configs/felhom-os-apply | the ONE table of root-owned paths + the installer of them | ANY new root-owned file the installer writes (sudoers line, wrapper, unit) — add it to the table, never a new installer fetch (R-840) | The builder (`scripts/build-config-bundle.py`) and the installer read the same table; `test_every_root_file_the_installer_writes_is_in_the_bundle` fails on a path the bundle lacks. Trust files (`/etc/felhom/os-trust.json`, `operator-signers`) are NEVER bundle paths (R17) |
 | `osupdate.ConfigUpdateExecutor` | internal/osupdate/bundle.go | signed op `agent_config_update` {agent_version, bundle_sha256} | delivering the bundle to an installed box | a courier only: the root wrapper re-verifies signature, host, nonce and sha itself |
 | `osupdate.Leg.SendUnsent` / `lockPass` (v0.144.0, R-868) | internal/osupdate/unsent.go | `(ctx) int` | an OS-pass report the agent never sent (killed mid-pass): the wrapper keeps `report-<run>-<layer>-apply.json` beside the plan; the agent deletes it once the hub has it | any new caller that runs an apply pass must hold `lockPass` (flock, across processes) — the sender must never take a running pass's copy |
 | `osupdate.LoadSavedBlock` (v0.144.0, R-866) | internal/osupdate/leg.go | `(planDir) (block, savedAt, ok)` | the hub's newest os_update block as the daemon last received it (`os-update-block.json`) | the debug pass uses it ONLY when the hub cannot be reached, and says so in its header; no saved block → no pass |
-| `guesthook.InstallSnippet` / `Register` | internal/guesthook/install.go | `InstallSnippet(ctx, runner) error` | pre-start self-heal hook install (C1 net) | Same random-temp+install pattern; snippet delegates to the agent binary (no shell logic). Issues `mkdir -p /var/lib/vz/snippets` FIRST (v0.63.0, B2 — fresh boxes lack the dir; sudoers grants exactly that argv) |
+| `dpkg_state()` / `DPKG_STATE_SCRIPT` (v0.145.0, R-876) | configs/felhom-os-apply | `audit, journal = self.dpkg_state()` | dpkg's state in ONE call: `--audit` AND the update journal | never gate a repair on `--audit` alone — a crash leaves only the journal (measured); keep it one call (R-845) |
+| `guesthook.SnippetReady` / `Register` (v0.146.0, R-861) | internal/guesthook/install.go | `SnippetReady(path) error` | is the pre-start hook (a FIXED file from the bundle) in place — register only then | The agent never installs the hook (Proxmox runs it as root); a missing hookscript stops a guest start, so never `Register` without `SnippetReady` |
 
 ### Disk / format safety (role gates, durable IDs, format guards)
 
@@ -63,8 +66,8 @@
 |---|---|---|---|---|
 | `IntentStore` (`Get/SetEnrolled/SetEjected/SetDecommissioned/OnAbsent`) | internal/storage/intent.go | `OpenIntentStore(path)` | drive intent (4-state self-heal) | Keyed by durable-id only; `OnAbsent` is the ONLY ejected→enrolled path; refuses empty ids |
 | `GuestBindStore` (`Record/Remove/Guests`) | internal/localapi/guestbindstore.go | `OpenGuestBindStore(path)` | per-guest enrolled binds (F9 re-assert) | Same tmp+rename 0600 pattern as IntentStore |
-| `FormatJobStore` + `startFormatDetached` + `RecoverFormatJob` | internal/localapi/formatjob.go | `startFormatDetached(device, durableID, fstype, blank) <-chan error` | detached, restart-surviving mkfs (F20-BUG3) | Runs off `s.baseCtx` (60-min bound) so a request deadline can't SIGKILL mkfs; recovery re-resolves by durable id; blank jobs re-check STILL-blank |
-| `TokenStore.Mint` / `Lookup` | internal/localapi/tokenstore.go | `Mint(vmid) (plaintext, error)` | per-guest local-API tokens | Only the SHA-256 hash persists (fsync'd append log); constant-time compare on lookup; plaintext returned exactly once. Lookup RELOADS the file once on a miss (v0.63.0, B3): the one-shot provisioner mints into the same file the daemon indexes — cross-process coherence without a restart; append-only size check bounds the re-read |
+| `FormatJobStore` + `startFormatDetached` + `RecoverFormatJob` | internal/localapi/formatjob.go | `startFormatDetached(device, durableID, fstype, blank) (*formatJob, <-chan error)` | detached, restart-surviving mkfs (F20-BUG3) | Runs off `s.baseCtx` (60-min bound) so a request deadline can't SIGKILL mkfs; recovery re-resolves by durable id; blank jobs re-check STILL-blank; on success `job.FSUUID` = the new filesystem's UUID, read back only when the durable id still resolves to the formatted device (R-25) — read it only after `done` delivers |
+| `TokenStore.Mint` / `Lookup` | internal/localapi/tokenstore.go | `Mint(vmid) (plaintext, error)` | per-guest local-API tokens | Only the SHA-256 hash persists (fsync'd append log); constant-time compare on lookup; plaintext returned exactly once. Lookup stats the file on EVERY call and reloads BEFORE answering when the append-only log grew (R-269; was reload-on-miss only, v0.63.0 B3, which let a token rotated out by another process keep authorizing as a map hit): the one-shot provisioner mints into the same file the daemon indexes — cross-process coherence both ways without a restart; unchanged size = no re-read. Pinned by `TestTokenStore_RotatedOutTokenRejectedFirst` |
 | `FileNonceStore.SeenOrRecord` | internal/authz/noncestore.go | `SeenOrRecord(nonce, exp) bool` | durable anti-replay | fsync'd before returning false; prune only after exp |
 | `Journal` (`Append/Latest/InFlight/AlreadyApplied`) | internal/reconcile/journal.go | `OpenJournal(path)` | op journal + idempotency + crash recovery | `Recover` consumes `InFlight()`; scratch entries special-cased |
 
@@ -118,7 +121,7 @@
 | Anti-retarget durable-id binding | internal/localapi/wipe_reresolve.go | resolve id → re-derive + exact match → re-inspect expected state → act on RE-RESOLVED device only |
 | Atomic single-file JSON store | internal/storage/intent.go | `Open*` loads (missing=empty, corrupt=fail-loud), mutex, tmp+rename 0600, idempotent set |
 | Durable append-only log + index | internal/authz/noncestore.go (`FileNonceStore`) | fsync before returning "new"; replay into index on open; expiry-only compaction |
-| Injectable seam funcs on Server | internal/localapi/server.go (`reresolveWipe`, `deviceDurableID`, `boundCheck`, `deviceCheck`, `livenessCheck`, net-verify: `netTrigger`/`netMounted`/`netJournal`/`netReachable`) | prod default wired in `NewServer`; tests override — no real /dev, /proc/mounts, journalctl or TCP in tests. **For mount-table predicates prefer the DATA seams `procSelfMountinfo` / `procGuestMountinfo` (internal/localapi/intermediary.go) over `boundCheck`/`livenessCheck`**: pointing them at a captured fixture runs the real parser, the real predicate and the real handler, so the test cannot go hollow the way R-116's did |
+| Injectable seam funcs on Server | internal/localapi/server.go (`reresolveWipe`, `deviceDurableID`, `boundCheck`, `deviceCheck`, `livenessCheck`, net-verify: `netTrigger`/`netMounted`/`netJournal`/`netReachable`; R-856 `crashGuardStatePath` — GET /host/crash-guard's state file, internal/localapi/crashguard.go) | prod default wired in `NewServer`; tests override — no real /dev, /proc/mounts, journalctl or TCP in tests. **For mount-table predicates prefer the DATA seams `procSelfMountinfo` / `procGuestMountinfo` (internal/localapi/intermediary.go) over `boundCheck`/`livenessCheck`**: pointing them at a captured fixture runs the real parser, the real predicate and the real handler, so the test cannot go hollow the way R-116's did |
 | `Server.devicePresent` (R-113, v0.114.0) | internal/localapi/disks.go | `devicePresent(rawMountPath) bool`; seam `deviceCheck`, default `isHostMountpoint` | the agent's DEVICE-presence signal — asks whether the drive's RAW mount is still mounted | **Use this, never the bind, to answer "is the drive there".** The raw mount is a device-bound systemd unit and dies with its device; the agent's own bind under the shared parent is NOT device-bound and outlives it as a stale shell. `BoundUnderParent` is now `boundUnderParent(...) && devicePresent(...)` at BOTH /disks construction sites — dropping either half is a regression with its own red-proof. Empty path ⇒ **true** (unknown is never absent: absent stops a customer's apps) |
 | `bindLiveness` + `BindLiveness` (R-117, v0.117.0) | internal/localapi/intermediary.go | `bindLiveness(stable, raw) BindLiveness`; seam `livenessCheck`; read verdicts ONLY via `.Usable()` | the agent's bind-LIVENESS signal — the third term of `BoundUnderParent` | **`devicePresent` and `boundUnderParent` are both PATH-PRESENCE tests and neither is liveness.** They compare only mountinfo field 5, so both stay true over a bind that names the drive that went away while the raw mount healed onto the returning one (measured: raw 8:32 /dev/sdc, bind 8:16 /dev/sdb `shutdown`, EIO both ways, payload healthy). Two dead states, and a fix needs BOTH checks: devno mismatch (the detach/return case) AND the ext4 abort tokens `shutdown`/`emergency_ro` (the steady-state case, where the devnos AGREE because the device never left). **THREE states, never a bool** — `BindUnknown` must exist and `Usable()` treats it as PRESENT (absent stops a customer's apps). **Order matters:** compare devices first and read the abort flag off the RAW mount in the stale case — abort-first classifies the real return state as aborted and refuses the re-bind that repairs it. **NO BLOCK I/O, ever** (CLAUDE.md rule; a probe on a wedged device survives SIGKILL). 6 red-proofs |
 | `AttachDrive` repair ruling (R-117, v0.117.0) | internal/localapi/intermediary.go | the `switch bindLiveness(...)` inside the `n == 1 && GuestSeesMount` arm | decides whether the existing self-heal runs | `BindStaleDevice` ⇒ **re-bind** (the raw mount is a healthy new superblock; repairs live, no guest restart). `BindAborted` ⇒ **quiet no-op** — a re-bind lands on the SAME dead superblock and this runs every 20 s, so re-binding is an infinite silent retry that also masks the state; it must surface via `BoundUnderParent=false`. `BindLive`/`BindUnknown` ⇒ no-op, unchanged. **Do not return an error for the aborted case** — the reconcile loop would log a failure every 20 s |
@@ -154,10 +157,13 @@
 | `storage.HostOps` | internal/storage/hostops.go | `*SudoHostOps` (prod), `NoopHostOps` (degraded) | fakes in internal/storage/observe_test.go, watchdog_test.go |
 | `storage.HostReader` | internal/storage/hostread.go | `*ProcHostReader` | `fakeHostReader` internal/localapi/disks_test.go; internal/storage/role_test.go. v0.87.0: `BlockSlaves(name)` lists `/sys/block/<name>/slaves` (root-free) — backs the `SystemDisks` dm/md walk (`physicalDisksOf`/`walkSlaves`, role.go); per-branch conservatism: an unresolvable slave fails the WHOLE walk → all-system fail-safe. NEVER weaken the signature test `TestSystemDisks_WalkTopologies` (root-backing disk always in the system set). |
 | `localapi.DiskOps` / `StorageGate` / `GuestAttacher` / `GuestLister` | internal/localapi/disks.go | `*storage.SudoHostOps`; `storageGateAdapter` (cmd/felhom-agent/main.go); `*GuestBinder`; `*proxmox.Client` | `fakeDiskOps`/`fakeGate`/`fakeGuestAttacher`/`fakeGuestList` internal/localapi/disks_test.go |
+| `lanresolver.hostRoot` + `dnsmasqUnitPaths` (data seam, R-317) | internal/lanresolver/lanresolver.go | prod `hostRoot = "/"`; probe = the `dnsmasq` package's systemd UNIT, never `/usr/sbin/dnsmasq` (owned by `dnsmasq-base`) | internal/lanresolver/ensure_dnsmasq_test.go — fixture root tree + recording `proxmox.Runner`; the REAL `os.Stat` probe and `EnsureDnsmasq` run. `TestEnsureDnsmasq_ProductionProbeIsTheUnit` pins the production wiring |
 | `localapi.GuestAPI` / `BackupService` / `BackupStore` / `TokenAuthority` | internal/localapi/server.go | `*proxmox.Client`, `*backup.BackupRunner`, `*backup.Store`, `*TokenStore` | `fakeGuests`/`fakeBackups`/`fakeStore` internal/localapi/server_test.go |
 | `backup.InFlight` | internal/backup/inflight.go | `TryAcquire(what) (release, busy, ok)` / `Busy()` | THE host-wide "one heavy guest operation at a time" gate — shared by the local-API backup path and the restore-test scheduler (R-85) | A **LINK** guard, not a lock one: the scratch VMID never touches the live guest's vzdump lock, but an offsite restore PULLS multi-GB over the tunnel a backup PUSHES one. Callers **DEFER, never cancel** — a deferred restore-test costs coverage, a cancelled backup costs the backup. A nil gate is ungated (pre-R-85 callers). |
+| `fstrim.Trimmer` (R-444) | internal/fstrim/fstrim.go | `New(runner, guests, gate, statePath, logger)` / `Pass(ctx)` / `GuestDiskTrimStatus(ctx)` / `ParseTrimmed(out)` | the weekly `pct fstrim <vmid>` of owned running guests (Wednesday from 10:00 local, starts 10:00-20:59 only), under `backup.InFlight`; last result per guest persisted and reported as `guest_disk_trim` | A busy gate DEFERS to the next hourly tick, never waits; a failed trim retries at most `MaxAttemptsPerWeek`; the report reads the persisted record, it never runs pct |
 | `capability` store-grant probe (`storeGrantStatuses` / `storeGrantVerdict` / `Client.Permissions`) | cmd/felhom-agent/main.go, internal/proxmox/query.go | *"may the agent READ this backup tier?"*, one `capability.Status` per configured tier | R-185. **Never infer permission from an empty content listing** — `{"data":[]}` is what a FORBIDDEN tier and a NEWBORN tier both return, and that ambiguity hid an unreadable host tier on both demo boxes. Ask `/access/permissions` **as the agent's own token** (root always says yes). **The ungranted answer is not empty and not a 403** — it carries the privileges inherited from the box-wide `/` grant, so test for **`Datastore.AllocateSpace`** specifically; path-presence or `Datastore.Audit` reports a blinded storage healthy. Probed set comes from `BackupTiers()`, never a fixed list. Critical except the `local` fallback. Composes AROUND the sudo prober (the `poolReadStatus` precedent); `Status`'s wire shape is untouched so the hub alert is free. Unreachable PVE ⇒ degraded, never ok. |
 | `backup.RestoreTestState` | internal/backup/restoretest_state.go | `RecordSuccess(target,archive,tier,verified,t)` / `ProvenArchive(target)` / `ProvenRestoreTests(ctx)` / `LastSuccess(target)` / `OldestFirst(targets)` | Per-tier restore-test PROOF state, persisted (atomic tmp+rename) — **which archive** was proven, and when (R-86) | **Credit ONLY on success** — a permanently failing tier must keep sorting first, or it looks freshly proven and stops being retried. Ties break on target id: without it, two tiers proven in the same second rotate by Go's randomised map order. **This one NEEDS persistence unlike R-84** — R-84 had ground truth to consult (the archive is still on the storage); a restore-test destroys its scratch and leaves no artifact. **R-86: the ARCHIVE is the state, the time is metadata** — a time alone cannot answer "have we proven THIS archive", which is the due-check's whole question. A pre-R-86 file (bare RFC3339 per target) keeps its time and yields NO proven archive, so each tier is due once after the upgrade; reading a legacy time as proof of the current archive would invent a guarantee. **R-189: it is also the REPORTABLE half of the restore-test signal.** The in-memory `backup.Store` holds only this process's latest run, and under per-archive due-ness the agent will not re-test a proven archive — so a proof lost to a restart is not repeated for a whole archive generation (observed live: a passing 14.5 GB offsite restore reached no host-report). `ProvenRestoreTests` renders the stored proofs as `hub.RestoreTest` entries and the collector merges them; a record missing the archive or the tier is NOT emitted, because an unproven tier reading as proven is worse than the defect. **Only successes are stored, deliberately:** a success suppresses future work, a failure causes it. |
+| `backup.BackupSuccessState` | internal/backup/backup_state.go | `RecordBackupSuccess(target, b)` / `LastKnownSuccess(target, vmid)` | Newest SUCCESSFUL backup per tier+guest, persisted (atomic tmp+rename) — the due-check's fallback when the tier's storage cannot be read after a restart (R-894) | **Read ONLY when the storage cannot be read** — a storage that answers is the ground truth (R-84), and an archive absent there must make the tier due even when this file remembers one. A saved copy older than the cadence still reads due. Only successes are written. |
 | `hub.ProvenRestoreTestReporter` + `Collector.SetProvenRestoreTests` | internal/hub/collect.go | the DURABLE restore-test source, merged with the in-memory one | R-189. Merge rule: **one entry per tier, newest by `TestedAt` wins** — a fresh failure beats a stored success (the failure is the news, and it lives nowhere else), a stored success beats a stale in-memory entry after a restart, and a tier never appears twice (the hub would read two tests). An unparseable timestamp counts as OLDER, so a malformed entry cannot displace a good one. **The wiring is pinned by an AST test** — the method this replaced (`RestoreTestState.Snapshot`) carried a doc comment naming a host-report gauge and had no caller for weeks. |
 | `backup.SpecBuilder` / `backup.TierPicker` / `(*BackupRunner).PickSettledRestoreCandidateOn` | internal/backup/schedule.go, runner.go | `func(ctx,archive) RestoreTestSpec`; `func(ctx,target,notAfter) (archive,landed,error)` | The per-run restore-test spec + per-tier **settled** candidate lookup (R-85, widened by R-86) | The spec is built **PER RUN**, never frozen at construction — the pre-R-85 immediately-invoked value made the offsite tier unschedulable AND went stale on any config change. `SourceTier` comes from **the archive**, never the configured target (the v0.100.0 rule). A tier with no archive returns `("", zero, nil)` — **`""` is NOT an error**, or every fresh box looks broken for its first week. **R-86: `notAfter` is the settle cutoff** (zero = no cutoff, which is what keeps `PickRestoreCandidateOn` a one-line call into it), and the picker now skips entries failing `archivePlausiblyComplete` — under per-archive due-ness an incomplete phantom would be picked forever, fail forever, never earn proof, and make the tier due at EVERY evaluation. |
 | `localapi.BackupTier` + `normalizeBackupTiers` / `config.BackupConfig.BackupTiers` | internal/localapi/backup_tiers.go, internal/config/config.go | `normalizeBackupTiers(tiers, legacy, cadence) []BackupTier`; `BackupTiers() ([]BackupTier, []string)` | THE R-82 multi-tier resolution — one runner per tier, primary first | **The untargeted local-API contract is FROZEN**: no `?target=` ⇒ primary tier ⇒ pre-R-82 response BYTES (Target is `omitempty` and stays empty). Never default a missing cadence — reject it and log the warning at ERROR. Never share one retention knob between tiers. Jobs are keyed by (vmid,target). |

cmd/felhom-agent/fstrim_wiring_test.go

@@ -0,0 +1,53 @@
+package main
+
+import (
+	"go/ast"
+	"go/parser"
+	"go/token"
+	"testing"
+)
+
+// R-444: the weekly trim has the guestnet shape (component + reporter seam + goroutine), so its wiring is asserted
+// from the AST like TestMainWiresGuestNetWatchdog — a unit-green trim job that main.go never starts is the inert-seam
+// defect. It must also share the ONE heavy-op gate (heavyOps), or it could run beside a backup.
+func TestMainWiresGuestDiskTrim(t *testing.T) {
+	fset := token.NewFileSet()
+	f, err := parser.ParseFile(fset, "main.go", nil, 0)
+	if err != nil {
+		t.Fatalf("parse main.go: %v", err)
+	}
+	var constructedWithGate, reporterWired, started bool
+	ast.Inspect(f, func(n ast.Node) bool {
+		switch node := n.(type) {
+		case *ast.CallExpr:
+			if fn, ok := node.Fun.(*ast.SelectorExpr); ok {
+				switch fn.Sel.Name {
+				case "New":
+					if pkg, ok := fn.X.(*ast.Ident); ok && pkg.Name == "fstrim" && len(node.Args) >= 3 {
+						if id, ok := node.Args[2].(*ast.Ident); ok && id.Name == "heavyOps" {
+							constructedWithGate = true
+						}
+					}
+				case "SetGuestDiskTrimReporter":
+					reporterWired = true
+				}
+			}
+		case *ast.GoStmt:
+			if sel, ok := node.Call.Fun.(*ast.SelectorExpr); ok && sel.Sel.Name == "Run" {
+				if id, ok := sel.X.(*ast.Ident); ok && id.Name == "diskTrim" {
+					started = true
+				}
+			}
+		}
+		return true
+	})
+	if !constructedWithGate {
+		t.Error("main.go never calls fstrim.New(..., heavyOps, ...) — no trim job, or one outside the heavy-op gate")
+	}
+	if !reporterWired {
+		t.Error("main.go never calls collector.SetGuestDiskTrimReporter — the guest_disk_trim stanza never reaches the hub")
+	}
+	if !started {
+		t.Error("main.go never starts the trim job with `go diskTrim.Run(ctx)`")
+	}
+}

cmd/felhom-agent/main.go

@@ -22,6 +22,7 @@ import (
 	"os/exec"
 	"os/signal"
 	"path/filepath"
+	"regexp"
 	"strconv"
 	"strings"
 	"sync"
@@ -37,6 +38,7 @@ import (
 	"gitea.dooplex.hu/admin/felhom-agent/internal/escrow"
 	"gitea.dooplex.hu/admin/felhom-agent/internal/fasttick"
 	"gitea.dooplex.hu/admin/felhom-agent/internal/felhomsshd"
+	"gitea.dooplex.hu/admin/felhom-agent/internal/fstrim"
 	"gitea.dooplex.hu/admin/felhom-agent/internal/guesthook"
 	"gitea.dooplex.hu/admin/felhom-agent/internal/guestnet"
 	"gitea.dooplex.hu/admin/felhom-agent/internal/hub"
@@ -870,11 +872,10 @@ func runDaemon(cfg config.Config, logger *slog.Logger, logRing *applog.Ring) int
 		"interval_s", hcfg.PollSeconds) // hub key intentionally not logged
 	// R-868 (v0.144.0): an OS pass whose agent was killed kept its report on disk — send it now (a pass that starts
 	// first sends it itself; the pass lock keeps the two apart).
-	go func() {
-		if n := osLeg.SendUnsent(ctx); n > 0 {
-			logger.Info("osupdate: sent kept report(s) at start", "count", n)
-		}
-	}()
+	// v0.144.1: and again every 5 minutes — the wrapper of a killed pass can finish AFTER the restart (measured).
+	go osLeg.SendUnsentLoop(ctx, 5*time.Minute, func(n int) {
+		logger.Info("osupdate: sent kept report(s)", "count", n)
+	})
 
 	// Reconcile (slice 4) runs alongside the hub loop, sharing the per-guest queue
 	// (doc 03 §10). At slice 4 the desired-state provider is empty (no hub serving
@@ -1479,6 +1480,25 @@ func runDaemon(cfg config.Config, logger *slog.Logger, logRing *applog.Ring) int
 		}
 		go runJanitor(ctx, jd)
 	}
+	// R-444 (`09` §3 decision 139): the weekly guest disk trim — `pct fstrim <vmid>` of every owned, running guest,
+	// Wednesday from 10:00 local, daytime only, under the one-heavy-op gate. Not part of the errc fan-out: a trim job
+	// must never be able to bring the agent down.
+	if cfg.DiskTrim.Enabled() {
+		dtMode := proxmox.RunnerMode(cfg.Privileged.Mode)
+		if dtMode == "" {
+			dtMode = proxmox.RunnerSudo
+		}
+		dtRunner := &proxmox.ExecRunner{Mode: dtMode, SudoPath: cfg.Privileged.SudoPath}
+		dtGuests := localapi.NewStaleLockController(px, dtRunner, reconcile.DefaultPool, logger)
+		if dtGuests != nil {
+			diskTrim := fstrim.New(dtRunner, dtGuests, heavyOps,
+				filepath.Join(cfg.OOB.WithDefaults().StateDir, "guest-disk-trim.json"), logger)
+			collector.SetGuestDiskTrimReporter(diskTrim)
+			go diskTrim.Run(ctx)
+		}
+	} else {
+		logger.Info("fstrim: weekly guest disk trim disabled by config (disk_trim.disable)")
+	}
 	if lanLoop != nil {
 		lanServers = 1
 		go func() { errc <- lanLoop.Run(ctx) }()
@@ -1881,12 +1901,15 @@ func buildLocalAPIServer(cfg config.Config, px *proxmox.Client, store *backup.St
 		BackupTiers:    apiTiers, // R-82: primary first; untargeted endpoints act on the primary
 		InFlight:       inFlight, // R-85: shared with the restore-test scheduler (Scenario F)
 		Store:          store,
-		Storage:        observer,
-		DriveTargets:   driveTargets,                    // Impl-2a: registry+units drives for the /disks view (union w/ Observe storages)
-		Smart:          storage.NewSmartReader(hostOps), // v0.95.0 Fix B: SMART for the union-path drives
-		HostReader:     storage.NewProcHostReader(),     // Impl-2b: durableIDForMount raw-mount fallback + role gate
-		Tokens:         tokens,
-		BackupCadence:  cfg.Backup.BackupCadence(),
+		// R-894: the newest success per tier on disk — the due-check's fallback when the storage cannot be
+		// read right after a restart. Same state dir as restore-test-state.json.
+		LastKnownBackups: backup.NewBackupSuccessState(filepath.Join(cfg.OOB.WithDefaults().StateDir, "backup-success-state.json")),
+		Storage:          observer,
+		DriveTargets:     driveTargets,                    // Impl-2a: registry+units drives for the /disks view (union w/ Observe storages)
+		Smart:            storage.NewSmartReader(hostOps), // v0.95.0 Fix B: SMART for the union-path drives
+		HostReader:       storage.NewProcHostReader(),     // Impl-2b: durableIDForMount raw-mount fallback + role gate
+		Tokens:           tokens,
+		BackupCadence:    cfg.Backup.BackupCadence(),
 		// Disk management (slice 8C): the privileged host surface + the data-bearing wipe gate.
 		Disks:       hostOps,
 		DiskGate:    storageGateAdapter{gate: gate, hostID: cfg.Hub.HostID},
@@ -2698,6 +2721,9 @@ func (e *escrowCeremonyErr) Error() string { return e.err.Error() }
 // restic password auto-attach), Create (R + self-verified blob), wipe the staged secret, upload
 // when asked. It PRINTS NOTHING — the output-mode shells own every byte of stdout/stderr. R is
 // returned for the caller to surface exactly once; escrow.Create never logs it and neither do we.
+// pbsStorageIDRe is a PVE storage id (letters, digits, '-', '_', '.'; starts with a letter) — never a path (R-861).
+var pbsStorageIDRe = regexp.MustCompile(`^[A-Za-z][A-Za-z0-9_.-]{0,63}$`)
+
 func escrowCeremony(ctx context.Context, cfg config.Config, logger *slog.Logger, opts escrowCeremonyOpts) (escrowCeremonyOutcome, *escrowCeremonyErr) {
 	var out escrowCeremonyOutcome
 	storage := opts.storage
@@ -2707,6 +2733,16 @@ func escrowCeremony(ctx context.Context, cfg config.Config, logger *slog.Logger,
 	if storage == "" {
 		return out, &escrowCeremonyErr{kind: "usage", err: fmt.Errorf("selftest=escrow-create requires -storage <pbs-storage-id> (or escrow.pbs_storage_id)")}
 	}
+	// R-861 (v0.146.0): this runs as ROOT through FELHOM_ESCROW, but agent.json is owned by the agent user. So the
+	// paths a root run reads never come from it: the PVE secret dir and the WireGuard state dir are the fixed defaults,
+	// and the storage id is a plain PVE id (no slash, no dot-dot) — a crafted id or dir would read another root file.
+	if os.Geteuid() == 0 {
+		cfg.Backup.PBSSecretDir = ""
+		cfg.WGTunnel.StateDir = ""
+	}
+	if !pbsStorageIDRe.MatchString(storage) {
+		return out, &escrowCeremonyErr{kind: "usage", err: fmt.Errorf("selftest=escrow-create: storage id %q is not a plain PVE storage id", storage)}
+	}
 	out.Storage = storage
 	keyPath := cfg.Backup.PBSEncKeyPath(storage)
 	if _, err := os.Stat(keyPath); err != nil {

cmd/felhom-agent/r861_escrow_storage_id_test.go

@@ -0,0 +1,31 @@
+package main
+
+import (
+	"context"
+	"io"
+	"log/slog"
+	"testing"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/config"
+)
+
+// R-861 (agent v0.146.0): the root escrow ceremony builds a file path from the storage id; an id that is a path is
+// refused before anything is read. RED-PROOF: drop the pbsStorageIDRe check → the "../" ids reach the key stat and
+// come back as a "setup" error instead of "usage".
+func TestEscrowCeremony_StorageIDIsNeverAPath(t *testing.T) {
+	lg := slog.New(slog.NewTextHandler(io.Discard, nil))
+	for _, id := range []string{"../../../etc/shadow", "a/b", "/etc/pve/priv/x", ".hidden", ""} {
+		cfg := config.Default()
+		cfg.Escrow.PBSStorageID = "" // the flag decides here
+		_, e := escrowCeremony(context.Background(), cfg, lg, escrowCeremonyOpts{storage: id})
+		if e == nil || e.kind != "usage" {
+			t.Errorf("storage id %q was not refused as usage (got %+v)", id, e)
+		}
+	}
+	cfg := config.Default()
+	cfg.Backup.PBSSecretDir = t.TempDir()
+	_, e := escrowCeremony(context.Background(), cfg, lg, escrowCeremonyOpts{storage: "felhom-pbs"})
+	if e == nil || e.kind != "setup" {
+		t.Fatalf("control: a plain id must pass the check and fail later on the missing key (setup), got %+v", e)
+	}
+}

cmd/felhom-agent/r894_wiring_test.go

@@ -0,0 +1,74 @@
+package main
+
+import (
+	"go/ast"
+	"testing"
+)
+
+// R-894 — the on-disk backup record is WIRED on the daemon path (the built-but-never-wired class).
+// main → runDaemon → buildLocalAPIServer, and inside it the localapi.Options literal carries
+// LastKnownBackups built by backup.NewBackupSuccessState. An AST walk, not a string match, for the
+// reasons in escrow_recover_wiring_test.go.
+//
+// COMPANION RED-PROOF (observed): delete the `LastKnownBackups:` line from buildLocalAPIServer → this
+// fails with "localapi.Options in buildLocalAPIServer has no LastKnownBackups field". Restored.
+func TestR894_LastKnownBackupsIsWiredIntoTheDaemon(t *testing.T) {
+	_, f := parseMain(t)
+	if !callsWithin(f, "main")["runDaemon"] || !callsWithin(f, "runDaemon")["buildLocalAPIServer"] {
+		t.Fatal("main → runDaemon → buildLocalAPIServer is broken — the path this test asserts is not the live one")
+	}
+	var field, built bool
+	for _, d := range f.Decls {
+		fd, ok := d.(*ast.FuncDecl)
+		if !ok || fd.Name == nil || fd.Name.Name != "buildLocalAPIServer" || fd.Body == nil {
+			continue
+		}
+		ast.Inspect(fd.Body, func(n ast.Node) bool {
+			cl, ok := n.(*ast.CompositeLit)
+			if !ok {
+				return true
+			}
+			sel, ok := cl.Type.(*ast.SelectorExpr)
+			if !ok {
+				return true
+			}
+			if pkg, _ := sel.X.(*ast.Ident); pkg == nil || pkg.Name+"."+sel.Sel.Name != "localapi.Options" {
+				return true
+			}
+			for _, el := range cl.Elts {
+				kv, ok := el.(*ast.KeyValueExpr)
+				if !ok {
+					continue
+				}
+				if k, ok := kv.Key.(*ast.Ident); ok && k.Name == "LastKnownBackups" {
+					field = true
+					if callsIn(kv.Value)["backup.NewBackupSuccessState"] {
+						built = true
+					}
+				}
+			}
+			return true
+		})
+	}
+	if !field {
+		t.Fatal("localapi.Options in buildLocalAPIServer has no LastKnownBackups field")
+	}
+	if !built {
+		t.Fatal("LastKnownBackups is not built by backup.NewBackupSuccessState")
+	}
+}
+
+func callsIn(n ast.Node) map[string]bool {
+	out := map[string]bool{}
+	ast.Inspect(n, func(n ast.Node) bool {
+		if ce, ok := n.(*ast.CallExpr); ok {
+			if fn, ok := ce.Fun.(*ast.SelectorExpr); ok {
+				if x, ok := fn.X.(*ast.Ident); ok {
+					out[x.Name+"."+fn.Sel.Name] = true
+				}
+			}
+		}
+		return true
+	})
+	return out
+}

configs/felhom-agent.sudoers

@@ -1,30 +1,38 @@
-# felhom-agent sudoers allowlist — the NARROW host-root surface (slice 5 Phase B, doc 03 §3/§7).
+# felhom-agent sudoers allowlist — the NARROW host-root surface (slice 5 Phase B, doc 03 §3/§7; narrowed R-861).
 #
-# Install as a drop-in: /etc/sudoers.d/felhom-agent (mode 0440, root:root), validated with
-# `visudo -cf`. The agent runs as the non-root `felhom-agent` service user and shells out via
-# `sudo -n` with FIXED argument vectors (no shell). The fine-grained validation is done IN
-# the agent BEFORE exec (internal/storage/validate.go): UUIDs against a strict hex regex,
-# mount paths confined+traversal-checked, SMART devices whitelisted to raw disks, LVM names
-# charset-checked. These sudoers wildcards are the COARSE allowlist; the agent is the fine
-# gate, so a wildcard can never be abused by a value the agent didn't already validate.
+# Install as a drop-in: /etc/sudoers.d/felhom-agent (mode 0440, root:root), validated with `visudo -cf`. It rides the
+# signed config bundle (R-840). The agent runs as the non-root `felhom-agent` user and shells out via `sudo -n` with
+# FIXED argument vectors (no shell).
 #
-# Binary paths MUST match the agent config (privileged.systemctl/install/smartctl/lvs). Adjust
-# for your distro (Debian/PVE shown). A missing/declined entry degrades the agent with a
-# warning (SMART→UNKNOWN, mount→logged error), it does not crash.
+# R-861 (agent v0.146.0) — EXACT PATTERNS, NOT GLOBS. A sudoers `*` in the ARGUMENTS also matches spaces, so
+# `pct set [0-9]* -onboot 1` matched `pct set 100 --dev0 /dev/sda -onboot 1` (a raw host disk for the guest), and
+# `mount --bind /mnt/*/felhom-data /mnt/felhom-drives/*` matched a `..` path onto /etc. Every argument list that varies
+# is now a sudo regular expression (`^...$`, sudo >= 1.9.10; Debian 13 / PVE 9 ship 1.9.16): one value per slot, a
+# fixed character set, no `..`, no extra argument. Lines with no variable part stay literal. The patterns are pinned
+# by the capability manifest (every real call must match: TestManifestCoveredBySudoers) and by injection cases that
+# must NOT match (configs/test_sudoers_patterns.py, and live with `sudo -l -U felhom-agent` on the demo boxes).
+#
+# R-861 — NO FILE THE AGENT WROTE IS INSTALLED WHERE ROOT READS IT. The `install` lines are gone: a mount unit, a
+# dnsmasq drop-in, the WireGuard config and the OOB sshd config + key go through `felhom-priv-apply`, a root wrapper
+# from the bundle that checks the CONTENT against the agent's own renderers; the guest pre-start hook and the shared
+# drive parent are FIXED files that come with the bundle itself; the agent binary is replaced only by an
+# operator-signed agent_update that `felhom-os-apply` verifies as root (`felhom-selfupdate-guarded apply` is no longer
+# here). The two remaining root runs of agent code (FELHOM_ESCROW, the guest hook) therefore run only a signed binary.
+#
+# Binary paths MUST match the agent config (privileged.systemctl/install/smartctl/lvs). A missing/declined entry
+# degrades the agent with a warning (SMART→UNKNOWN, mount→logged error), it does not crash; the capability probe
+# reports it to the hub.
 
 Cmnd_Alias FELHOM_MOUNT = \
-  /usr/bin/install -o root -g root -m 0644 -- /var/lib/felhom-agent/units/* /etc/systemd/system/*.mount, \
+  /usr/local/sbin/felhom-priv-apply ^unit mnt-[A-Za-z0-9_.\\-]+\.(mount|automount)$, \
   /usr/bin/systemctl daemon-reload, \
-  /usr/bin/systemctl enable --now -- *.mount, \
-  /usr/bin/systemctl disable -- *.mount, \
-  /usr/bin/systemctl stop -- *.mount
+  /usr/bin/systemctl ^enable --now -- mnt-[A-Za-z0-9_.\\-]+\.mount$, \
+  /usr/bin/systemctl ^disable -- mnt-[A-Za-z0-9_.\\-]+\.mount$, \
+  /usr/bin/systemctl ^stop -- mnt-[A-Za-z0-9_.\\-]+\.mount$
 
 Cmnd_Alias FELHOM_DISK = \
-  /usr/sbin/smartctl -a -j /dev/sd[a-z]*, \
-  /usr/sbin/smartctl -a -j /dev/nvme[0-9]*n[0-9]*, \
-  /usr/sbin/smartctl -a -j /dev/vd[a-z]*, \
-  /usr/sbin/smartctl -a -j /dev/hd[a-z]*, \
-  /usr/sbin/lvs --reportformat json --units b -o lv_name\,data_percent\,metadata_percent -- *, \
+  /usr/sbin/smartctl ^-a -j /dev/(sd[a-z]+|nvme[0-9]+n[0-9]+|vd[a-z]+|hd[a-z]+)$, \
+  /usr/sbin/lvs ^--reportformat json --units b -o lv_name\,data_percent\,metadata_percent -- [A-Za-z0-9_.+-]+(/[A-Za-z0-9_.+-]+)?$, \
   /usr/sbin/pvs --reportformat json --noheadings -o pv_name, \
   /usr/sbin/zpool status -P
 
@@ -35,9 +43,9 @@ Cmnd_Alias FELHOM_DISK = \
 # (the wildcard only ever names a path the agent itself created), and the bootstrap file the agent
 # writes there is the only thing these touch. ':' is escaped per sudoers grammar.
 Cmnd_Alias FELHOM_PROVISION = \
-  /usr/bin/chown -R 100000\:100000 /var/lib/felhom-agent/guests/*, \
-  /usr/sbin/pct set [0-9]* -mp[0-9]* /var/lib/felhom-agent/guests/*, \
-  /usr/sbin/pct set [0-9]* -onboot 1
+  /usr/bin/chown ^-R 100000\:100000 /var/lib/felhom-agent/guests/[0-9]+(/bootstrap)?$, \
+  /usr/sbin/pct ^set [0-9]+ -mp[0-9]+ /var/lib/felhom-agent/guests/[0-9]+/bootstrap\,mp\=/[A-Za-z0-9/_.-]+(\,ro\=1)?$, \
+  /usr/sbin/pct ^set [0-9]+ -onboot 1$
 
 # Disk inspection + format (slice 8C + Impl-1). blkid/lsblk read the device's data-bearing evidence
 # (the agent decides data-bearing-ness from THIS, never the caller's claim). Format goes ONLY through
@@ -45,66 +53,54 @@ Cmnd_Alias FELHOM_PROVISION = \
 # mkfs the OS disk — the wrapper re-checks the catastrophic cases (system disk / LVM PV / foreign mount)
 # as root and refuses, and the agent's unclaimed-disk filter (claim.go) is the primary guard above it.
 Cmnd_Alias FELHOM_FORMAT = \
-  /usr/sbin/blkid -p -o export /dev/*, \
-  /usr/bin/lsblk -J -o NAME\,FSTYPE\,PTTYPE\,MOUNTPOINT /dev/*, \
-  /usr/local/sbin/felhom-mkfs-guarded /dev/* *
+  /usr/sbin/blkid ^-p -o export /dev/[^ ]+$, \
+  /usr/bin/lsblk ^-J -o NAME\,FSTYPE\,PTTYPE\,MOUNTPOINT /dev/[^ ]+$, \
+  /usr/local/sbin/felhom-mkfs-guarded ^/dev/[^ ]+ (ext4|xfs)$
 
 # LAN split-horizon resolver (internal/lanresolver): the agent manages a host-side dnsmasq that
-# answers *.<customer-domain> with each guest's live LAN IP. install only ever writes felhom-*.conf
-# drop-ins (from agent-written /tmp temp files); the two `pct exec` reads are FIXED command vectors
+# answers *.<customer-domain> with each guest's live LAN IP. A felhom-*.conf drop-in reaches /etc/dnsmasq.d
+# only through felhom-priv-apply, which allows exactly the lines the resolver renders (R-861: a `dhcp-script=` would
+# run as root); the two `pct exec` reads are FIXED command vectors
 # (the guest's eth0 IPv4 + the controller's pulled controller.yaml for the domain) — NOT a general
 # `pct exec`. systemctl is scoped to the dnsmasq unit only. The agent never edits /etc/resolv.conf.
 Cmnd_Alias FELHOM_DNSMASQ = \
   /usr/bin/apt-get install -y -q dnsmasq, \
-  /usr/bin/install -m 0644 /tmp/felhom-resolver-*.conf /etc/dnsmasq.d/felhom-*.conf, \
+  /usr/local/sbin/felhom-priv-apply ^dnsmasq /tmp/felhom-resolver-[0-9]+\.conf felhom-[a-z0-9][a-z0-9._-]*\.conf$, \
   /usr/bin/systemctl enable --now dnsmasq, \
   /usr/bin/systemctl reload dnsmasq, \
   /usr/bin/systemctl restart dnsmasq, \
-  /usr/bin/rm -f /etc/dnsmasq.d/felhom-*.conf, \
-  /usr/sbin/pct exec [0-9]* -- ip -4 -o addr show dev eth0, \
-  /usr/sbin/pct exec [0-9]* -- docker exec felhom-controller cat /opt/docker/felhom-controller/controller.yaml
+  /usr/bin/rm ^-f /etc/dnsmasq\.d/felhom-[a-z0-9][a-z0-9._-]*\.conf$, \
+  /usr/sbin/pct ^exec [0-9]+ -- ip -4 -o addr show dev eth0$, \
+  /usr/sbin/pct ^exec [0-9]+ -- docker exec felhom-controller cat /opt/docker/felhom-controller/controller\.yaml$
 
-# Guest mountpoint lifecycle (intermediary-mount re-architecture + C1 net). The pre-start self-heal hook
-# wrapper is installed once into the PVE snippets dir (from an agent-written /tmp file) and registered
-# per-guest; decommission/eject DELETE the dead mountpoint slot so a missing bind source can't brick the
-# guest at next boot (the B3 C1 fix). The agent fine-validates the vmid (numeric) + slot (mp[0-9]+) and
-# the snippet path is fixed — the wildcards are the coarse allowlist. The install SOURCE is a
-# random-named agent temp (os.CreateTemp, audit B1 — a fixed /tmp name was a local TOCTOU), hence the
-# glob; the DESTINATION stays pinned. The `mkdir -p` creates the snippets dir on a FRESH box —
-# `install` won't create parents, so without it the hook install failed silently on Day-0 boxes
-# (B2, DRILL-day0-cleanroom-2026-07-03; fixed agent v0.63.0).
+# Guest mountpoint lifecycle (intermediary-mount re-architecture + C1 net). The pre-start self-heal hook is a FIXED file
+# from the config bundle (/var/lib/vz/snippets/felhom-guest-hook.sh — R-861: the agent no longer installs it from /tmp;
+# Proxmox runs it as root at every guest start). The agent only registers it per guest, deletes a dead mountpoint slot
+# (the B3 C1 fix) and reboots a guest to activate binds — each with an exact vmid / slot.
 Cmnd_Alias FELHOM_GUESTHOOK = \
-  /usr/bin/mkdir -p /var/lib/vz/snippets, \
-  /usr/bin/install -m 0755 -- /tmp/felhom-guest-hook-*.sh /var/lib/vz/snippets/felhom-guest-hook.sh, \
-  /usr/sbin/pct set [0-9]* --hookscript local\:snippets/felhom-guest-hook.sh, \
-  /usr/sbin/pct set [0-9]* --delete mp[0-9]*, \
-  /usr/sbin/pct reboot [0-9]*
+  /usr/sbin/pct ^set [0-9]+ --hookscript local\:snippets/felhom-guest-hook\.sh$, \
+  /usr/sbin/pct ^set [0-9]+ --delete mp[0-9]+$, \
+  /usr/sbin/pct ^reboot [0-9]+$
 
 # Intermediary mount model (the drive hot-swap re-architecture). The agent keeps a SHARED host parent
-# /mnt/felhom-drives (self-bind + make-shared + a boot-persistence systemd unit) and binds/unbinds each
-# drive's felhom-data namespace UNDERNEATH it so the change propagates into the running guest live (no
-# pct, no reboot). The agent fine-validates the drive name + confines paths before any exec; the trailing
-# `*` (matching the comma-laden mp spec) mirrors the existing FELHOM_PROVISION pattern.
-# `lxc-info -n <vmid> -p -H` resolves the guest init PID for the GuestSeesMount / bound_under_parent check
-# (a READ — the drive-gate's "is the drive live in the guest?" signal); WITHOUT it the non-root agent gets
-# an empty PID and reports every drive absent (multi-drive flapping, audit 2026-06-29). `make-private`
-# isolates the parent's peer group on FIRST setup only (EnsureSharedParent guards on mountpoint, so it
-# never re-churns a live parent); without it the parent stays in root's group and submounts double.
+# /mnt/felhom-drives (self-bind + make-shared) and binds/unbinds each drive's felhom-data namespace UNDERNEATH it so the
+# change propagates into the running guest live. The boot-persistence script + unit are FIXED files from the config
+# bundle (R-861: the agent no longer installs them from /tmp); the agent only enables the unit. A drive name is one
+# path segment that cannot start with a dot (no `..`); `lxc-info -n <vmid> -p -H` resolves the guest init PID for the
+# GuestSeesMount check; `make-private` isolates the parent's peer group on FIRST setup only.
 Cmnd_Alias FELHOM_INTERMEDIARY = \
   /usr/bin/mkdir -p /mnt/felhom-drives, \
-  /usr/bin/mkdir -p /mnt/felhom-drives/*, \
-  /usr/bin/mkdir -p /mnt/*/felhom-data, \
-  /usr/bin/chown 100000\:100000 /mnt/*/felhom-data, \
+  /usr/bin/mkdir ^-p /mnt/felhom-drives/[A-Za-z0-9_-][A-Za-z0-9_.-]*$, \
+  /usr/bin/mkdir ^-p /mnt/[A-Za-z0-9_-][A-Za-z0-9_.-]*/felhom-data$, \
+  /usr/bin/chown ^100000\:100000 /mnt/[A-Za-z0-9_-][A-Za-z0-9_.-]*/felhom-data$, \
   /usr/bin/mount --bind /mnt/felhom-drives /mnt/felhom-drives, \
   /usr/bin/mount --make-shared /mnt/felhom-drives, \
   /usr/bin/mount --make-private /mnt/felhom-drives, \
-  /usr/bin/mount --bind /mnt/*/felhom-data /mnt/felhom-drives/*, \
-  /usr/bin/umount /mnt/felhom-drives/*, \
-  /usr/bin/install -m 0755 -- /tmp/felhom-shared-parent-*.sh /usr/local/sbin/felhom-shared-parent.sh, \
-  /usr/bin/install -m 0644 -- /tmp/felhom-shared-parent-*.service /etc/systemd/system/felhom-shared-parent.service, \
+  /usr/bin/mount ^--bind /mnt/[A-Za-z0-9_-][A-Za-z0-9_.-]*/felhom-data /mnt/felhom-drives/[A-Za-z0-9_-][A-Za-z0-9_.-]*$, \
+  /usr/bin/umount ^/mnt/felhom-drives/[A-Za-z0-9_-][A-Za-z0-9_.-]*$, \
   /usr/bin/systemctl enable felhom-shared-parent.service, \
-  /usr/bin/lxc-info -n [0-9]* -p -H, \
-  /usr/sbin/pct set [0-9]* -mp8 /mnt/felhom-drives*
+  /usr/bin/lxc-info ^-n [0-9]+ -p -H$, \
+  /usr/sbin/pct ^set [0-9]+ -mp8 /mnt/felhom-drives\,mp\=/mnt/felhom-drives$
 
 # Controller-swap / managed auto-update (Option A, non-root). The agent owns the in-guest controller
 # image SWAP (it survives the controller being killed mid-swap): read the baked image ref, check the
@@ -119,11 +115,11 @@ Cmnd_Alias FELHOM_INTERMEDIARY = \
 #                                 the agent strict-validates the ref (controllerImageRe) before the write.
 # Validated GO: felhom.eu/documentation/audits/SPIKE-controllerswap-narrow-grants-2026-06-29.md.
 Cmnd_Alias FELHOM_CONTROLLERSWAP = \
-  /usr/sbin/pct exec [0-9]* -- cat /etc/felhom-controller-image, \
-  /usr/sbin/pct exec [0-9]* -- docker image inspect *, \
-  /usr/sbin/pct exec [0-9]* -- docker inspect -f *, \
-  /usr/sbin/pct exec [0-9]* -- systemctl restart felhom-controller-bootstrap.service, \
-  /usr/sbin/pct exec [0-9]* -- tee /etc/felhom-controller-image
+  /usr/sbin/pct ^exec [0-9]+ -- cat /etc/felhom-controller-image$, \
+  /usr/sbin/pct ^exec [0-9]+ -- docker image inspect gitea\.dooplex\.hu/admin/felhom-controller\:[0-9]+\.[0-9]+\.[0-9]+$, \
+  /usr/sbin/pct ^exec [0-9]+ -- docker inspect -f .+ (felhom-controller|cloudflared)$, \
+  /usr/sbin/pct ^exec [0-9]+ -- systemctl restart felhom-controller-bootstrap\.service$, \
+  /usr/sbin/pct ^exec [0-9]+ -- tee /etc/felhom-controller-image$
 
 # Stale-lock recovery (F2-b, v0.49.0). A host reboot DURING a vzdump backup leaves the guest with a
 # `snapshot-delete`/`backup` lock + `onboot:1` then can't start it → the customer box stays DOWN. The
@@ -131,7 +127,15 @@ Cmnd_Alias FELHOM_CONTROLLERSWAP = \
 # with no API equivalent (snapshot-delete + start go through the API token); the agent fine-validates the
 # vmid (numeric) before exec — the `[0-9]*` is the coarse allowlist.
 Cmnd_Alias FELHOM_STALELOCK = \
-  /usr/sbin/pct unlock [0-9]*
+  /usr/sbin/pct ^unlock [0-9]+$
+
+# Weekly guest disk trim (R-444, operator ruling `09` §3 decision 139). A thin pool only ever grows from blocks the
+# guest has FREED: `fstrim` inside the unprivileged container is refused (FITRIM: Operation not permitted), so the host
+# trims the guest's mounts. Measured on demo-hp 2026-10-06: `pct fstrim 9201` rc 0 in 24.4 s, pool 65.53 % -> 33.40 %,
+# apps kept answering. ONE exact pattern: a vmid and nothing else — no `--ignore-mountpoints`, no second argument
+# (pinned: TestSudoersFstrimRuleIsExact). The agent runs it on a weekly daytime timer under the heavy-op gate.
+Cmnd_Alias FELHOM_FSTRIM = \
+  /usr/sbin/pct ^fstrim [0-9]+$
 
 # Restore-test scratch teardown (F-LEAK, Campaign 8, v0.110.0). A restore-test whose restore FAILS
 # leaves a scratch guest the API token CANNOT destroy: `FelhomAgentGuest` is granted at /pool/felhom and
@@ -160,8 +164,9 @@ Cmnd_Alias FELHOM_SCRATCH_TEARDOWN = \
 # into the guest through the existing shared bind (an unprivileged LXC cannot mount NFS/CIFS itself).
 # A NAS is NOT a drive — no durable-id, no SMART, no wipe; these grants only install/enable/remove the
 # unit pair. The agent fine-validates every value (share name, server, export, uid/gid, creds path) before
-# any unit is rendered (internal/storage/netmount.go ValidateNetworkMountSpec); the trailing globs are the
-# COARSE allowlist. The `.mount` install/enable/disable/stop reuse FELHOM_MOUNT; this alias adds the
+# any unit is rendered (internal/storage/netmount.go ValidateNetworkMountSpec); the unit FILE reaches
+# /etc/systemd/system only through `felhom-priv-apply unit` (FELHOM_MOUNT), which requires nosuid,nodev on a network
+# share (R-861). The `.mount` enable/disable/stop reuse FELHOM_MOUNT; this alias adds the
 # `.automount` variants + the unit-file removal. The unit FILE name is the systemd-escaped mountpoint,
 # which always begins `mnt-felhom` (the mountpoint is /mnt/felhom-drives/<name>), so the rm glob is scoped
 # to felhom mount units only. mkdir of the mountpoint reuses FELHOM_INTERMEDIARY's /mnt/felhom-drives/*.
@@ -176,51 +181,46 @@ Cmnd_Alias FELHOM_SCRATCH_TEARDOWN = \
 #    behind (the campaign accumulated 10 stub-shaped leftovers). rmdir ONLY (never rm -rf): it refuses
 #    a non-empty dir, so unexpected data is preserved, not destroyed — a fail-safe grant.
 Cmnd_Alias FELHOM_NETMOUNT = \
-  /usr/bin/install -o root -g root -m 0644 -- /var/lib/felhom-agent/units/* /etc/systemd/system/*.automount, \
-  /usr/bin/systemctl enable --now -- *.automount, \
-  /usr/bin/systemctl disable -- *.automount, \
-  /usr/bin/systemctl stop -- *.automount, \
-  /usr/bin/systemctl reset-failed -- mnt-felhom*, \
-  /usr/bin/rmdir /mnt/felhom-drives/*, \
-  /usr/bin/rm -f /etc/systemd/system/mnt-felhom*
+  /usr/bin/systemctl ^enable --now -- mnt-[A-Za-z0-9_.\\-]+\.automount$, \
+  /usr/bin/systemctl ^disable -- mnt-[A-Za-z0-9_.\\-]+\.automount$, \
+  /usr/bin/systemctl ^stop -- mnt-[A-Za-z0-9_.\\-]+\.automount$, \
+  /usr/bin/systemctl ^reset-failed -- mnt-felhom[A-Za-z0-9_.\\-]*\.(mount|automount)$, \
+  /usr/bin/rmdir ^/mnt/felhom-drives/[A-Za-z0-9_-][A-Za-z0-9_.-]*$, \
+  /usr/bin/rm ^-f /etc/systemd/system/mnt-felhom[A-Za-z0-9_.\\-]*\.(mount|automount)$
 
 # Offsite WG tunnel (S3, doc 06 §3.3). The agent manages wg-quick@wg-felhom as an agent-managed
 # host service (the dnsmasq/lanresolver shape): conf staged in the agent-owned StateDir (never
-# /tmp), installed 0600 to the FIXED destination, unit enable/restart/disable. The ONLY wg read
+# /tmp), installed 0600 to the FIXED destination by `felhom-priv-apply wg`, which refuses any key renderConf never
+# writes (R-861: PostUp/PreUp run as root under wg-quick), unit enable/restart/disable. The ONLY wg read
 # is `latest-handshakes` — `wg show <if> dump` is FORBIDDEN everywhere (its interface line
-# carries the PRIVATE KEY; the S1 session-log incident). Both install paths are FIXED (no glob):
-# the agent has exactly one tunnel conf to manage.
+# carries the PRIVATE KEY; the S1 session-log incident). Source and destination are fixed in the wrapper.
 Cmnd_Alias FELHOM_WG = \
   /usr/bin/apt-get install -y -q wireguard-tools, \
-  /usr/bin/install -o root -g root -m 0600 -- /var/lib/felhom-agent/wg/wg-felhom.conf /etc/wireguard/wg-felhom.conf, \
+  /usr/local/sbin/felhom-priv-apply wg, \
   /usr/bin/systemctl enable --now wg-quick@wg-felhom, \
   /usr/bin/systemctl restart wg-quick@wg-felhom, \
   /usr/bin/systemctl disable --now wg-quick@wg-felhom, \
   /usr/bin/wg show wg-felhom latest-handshakes
 
-# Agent self-update (TASK D1, SPIKE-agent-selfupdate-2026-07-05). The agent downloads the
-# operator-SIGNED binary (sha256 pinned in the signed op — neither hub nor Gitea compromise can
-# substitute it), verifies the sha in-process, then hands off to the guarded wrapper, which
-# RE-verifies the sha as root, confines the staged path to /var/lib/felhom-agent/selfupdate/,
-# performs the A/B flip (atomic same-fs rename, .prev retained) and schedules a detached restart.
-# The apply args are a COARSE glob (spike S4b: sudoers fnmatch makes a [a-f0-9]* sha pattern
-# first-char-only anyway) — the wrapper's own sha re-verify + path confinement is the real gate.
-# `rollback` is normally run by felhom-agent-rollback.service (root, OnFailure=), not via sudo;
-# granting it here keeps the verb probe-able (capability self-check) and operator-invokable.
+# Agent self-update (TASK D1; R-861). The A/B flip (`felhom-selfupdate-guarded apply`) is NO LONGER the agent's: the
+# agent hands the operator-SIGNED agent_update to felhom-os-apply (FELHOM_OSAPPLY, mode agent_update), which verifies
+# the signature as root and only then runs the flip. Until v0.146.0 the agent passed the sha itself, so a compromised
+# agent could install any binary — the binary FELHOM_ESCROW and the guest hook run as root. `commit` (clear the pending
+# marker) and `rollback` (pending-guarded revert, normally run by felhom-agent-rollback.service) stay.
 Cmnd_Alias FELHOM_SELFUPDATE = \
-  /usr/local/sbin/felhom-selfupdate-guarded apply /var/lib/felhom-agent/selfupdate/* *, \
   /usr/local/sbin/felhom-selfupdate-guarded commit, \
   /usr/local/sbin/felhom-selfupdate-guarded rollback
 
 # Dedicated OOB sshd (TASK H1). The agent manages felhom-sshd like wg-felhom/dnsmasq: it RENDERS the
 # config (Port from its claim) + the operator's authorized_keys, validates with `sshd -t`, and reloads
-# (never restart-on-change [SF-2]). Both install SOURCES are the agent-owned staged files under
-# StateDir; both DESTINATIONS are FIXED. `sshd -t/-T` are the validate/discover reads. The
+# (never restart-on-change [SF-2]). Both files reach /etc/felhom-sshd only through felhom-priv-apply (R-861): the config
+# must be the ONE template with only the Port varying (an AuthorizedKeysFile the agent owns + `StrictModes no` would be
+# a root login), the key file one plain public key without options. `sshd -t/-T` are the validate/discover reads. The
 # systemctl verbs are SCOPED to felhom-sshd only. reset-failed precedes a deliberate restart [SF-5].
 # NOTHING here can touch the stock sshd, :22, or /etc/ssh.
 Cmnd_Alias FELHOM_SSHD = \
-  /usr/bin/install -o root -g root -m 0644 -- /var/lib/felhom-agent/felhom-sshd/sshd_config /etc/felhom-sshd/sshd_config, \
-  /usr/bin/install -o root -g root -m 0644 -- /var/lib/felhom-agent/felhom-sshd/authorized_keys.felhom-op /etc/felhom-sshd/authorized_keys/felhom-op, \
+  /usr/local/sbin/felhom-priv-apply sshd-config, \
+  /usr/local/sbin/felhom-priv-apply sshd-key, \
   /usr/sbin/sshd -t -f /var/lib/felhom-agent/felhom-sshd/sshd_config, \
   /usr/sbin/sshd -t -f /etc/felhom-sshd/sshd_config, \
   /usr/sbin/sshd -T -f /etc/felhom-sshd/sshd_config, \
@@ -270,8 +270,8 @@ Cmnd_Alias FELHOM_OOB = \
   /usr/sbin/nft list set inet felhom_oob ssh_port, \
   /usr/sbin/nft flush set inet felhom_oob operator_ips, \
   /usr/sbin/nft flush set inet felhom_oob ssh_port, \
-  /usr/sbin/nft add element inet felhom_oob operator_ips *, \
-  /usr/sbin/nft add element inet felhom_oob ssh_port *
+  /usr/sbin/nft ^add element inet felhom_oob operator_ips \{ [0-9.]+(/[0-9]+)? \}$, \
+  /usr/sbin/nft ^add element inet felhom_oob ssh_port \{ [0-9]+ \}$
 
 # Escrow ceremony (controller-driven, TASK 2026-07-13; mechanics validated by
 # SPIKE-controller-escrow-2026-07-13). ONE fixed argv — sudoers matches the argument vector
@@ -307,9 +307,9 @@ Cmnd_Alias FELHOM_OSAPPLY = \
   /usr/local/sbin/felhom-os-apply --plan /var/lib/felhom-agent/os/plan-*.json
 
 Cmnd_Alias FELHOM_GUESTNET = \
-  /usr/sbin/pct exec [0-9]* -- ip route show default, \
-  /usr/sbin/pct exec [0-9]* -- cat /etc/network/interfaces, \
-  /usr/sbin/pct exec [0-9]* -- pgrep -x dhclient, \
-  /usr/sbin/pct exec [0-9]* -- dhclient -pf /run/dhclient.eth0.pid -lf /var/lib/dhcp/dhclient.eth0.leases eth0
+  /usr/sbin/pct ^exec [0-9]+ -- ip route show default$, \
+  /usr/sbin/pct ^exec [0-9]+ -- cat /etc/network/interfaces$, \
+  /usr/sbin/pct ^exec [0-9]+ -- pgrep -x dhclient$, \
+  /usr/sbin/pct ^exec [0-9]+ -- dhclient -pf /run/dhclient\.eth0\.pid -lf /var/lib/dhcp/dhclient\.eth0\.leases eth0$
 
-felhom-agent ALL=(root) NOPASSWD: FELHOM_MOUNT, FELHOM_DISK, FELHOM_PROVISION, FELHOM_FORMAT, FELHOM_DNSMASQ, FELHOM_GUESTHOOK, FELHOM_INTERMEDIARY, FELHOM_CONTROLLERSWAP, FELHOM_STALELOCK, FELHOM_NETMOUNT, FELHOM_WG, FELHOM_SELFUPDATE, FELHOM_SSHD, FELHOM_OOB, FELHOM_PBSDR, FELHOM_BACKUPTARGET, FELHOM_SELFHEAL, FELHOM_ESCROW, FELHOM_GUESTNET, FELHOM_SCRATCH_TEARDOWN, FELHOM_OSAPPLY
+felhom-agent ALL=(root) NOPASSWD: FELHOM_MOUNT, FELHOM_DISK, FELHOM_PROVISION, FELHOM_FORMAT, FELHOM_DNSMASQ, FELHOM_GUESTHOOK, FELHOM_INTERMEDIARY, FELHOM_CONTROLLERSWAP, FELHOM_STALELOCK, FELHOM_FSTRIM, FELHOM_NETMOUNT, FELHOM_WG, FELHOM_SELFUPDATE, FELHOM_SSHD, FELHOM_OOB, FELHOM_PBSDR, FELHOM_BACKUPTARGET, FELHOM_SELFHEAL, FELHOM_ESCROW, FELHOM_GUESTNET, FELHOM_SCRATCH_TEARDOWN, FELHOM_OSAPPLY

configs/felhom-guest-hook.sh

@@ -0,0 +1,4 @@
+#!/bin/sh
+# felhom-agent guest pre-start self-heal hook (C1 net). PVE calls: <script> <vmid> <phase>.
+/usr/local/bin/felhom-agent guest-hook "$1" "$2" || true
+exit 0

configs/felhom-os-apply

@@ -32,6 +32,14 @@
 #               held packages, kernel taint, the crash guard; guest Debian, Docker engine, containerd, live-restore.
 #   live-restore-on  (v0.142.0, layer guest) the ONE-TIME step of `09` decision 87: merge `"live-restore": true`
 #               into the guest's /etc/docker/daemon.json and `systemctl reload docker`. NEVER a restart (R-835).
+#   oom-check   (R-528, `09` decision 157; layer docker, lane slow) ONLY the memory-kill check below: no apt, no engine
+#               change, no authority needed. Wrapper-only — the agent never writes this plan; run it by hand as root.
+# R-528 (`09` decision 157): a docker-layer APPLY also runs `oom_check()` after health_after and reports it as
+#   "oom_check": {"result": "pass"|"fail"|"error", "oom_killed": bool, "oom_event": bool, "exit_code": int|null,
+#                 "image": str|null, "detail": str}
+#   — one throwaway container (the controller's own image, no network / volume / port, 64 MB cap) is made to exceed its
+#   memory; "pass" only when the engine says OOMKilled=true AND emits the `oom` event. It never changes the step's
+#   outcome or health: the hub decides whether the engine set can be approved. Pinned by the OOMCheck tests.
 # Output: log lines on stderr and the journal (tag felhom-os-apply); the LAST stdout line is
 #   OSAPPLY-REPORT <one JSON object>
 # which is what the agent parses. Exit 0 = done; 2 = refused (nothing changed); 3 = failed during install.
@@ -63,8 +71,21 @@ SNAPSHOT_LIST = "/etc/apt/sources.list.d/felhom-os-snapshot.list"
 APT_ENV = ["env", "DEBIAN_FRONTEND=noninteractive", "APT_LISTCHANGES_FRONTEND=none", "NEEDRESTART_MODE=l", "LC_ALL=C"]
 DPKG_OPTS = ["-o", "Dpkg::Options::=--force-confold", "-o", "Dpkg::Options::=--force-confdef"]
 MIN_FREE = 500 * 1024 * 1024
+# R-876: dpkg's state in ONE call — `--audit`, a marker line, then the update journal's file names.
+JOURNAL_MARK = "@@FELHOM-DPKG-JOURNAL@@"
+DPKG_STATE_SCRIPT = "dpkg --audit; echo " + JOURNAL_MARK + "; ls -A /var/lib/dpkg/updates 2>/dev/null; true"
 # The installer's ROOT-OWNED record (felhom-host-install.sh `state_set mode`); the agent cannot write it.
 INSTALL_STATE = "/var/lib/felhom-install/state.json"
+# R-528: the memory-kill check (oom_check). One 200 MB block under a 64 MB cap: measured on Docker 29.8.2 to be
+# OOM-killed with OOMKilled=true and an `oom` event. Every call is bounded: timeouts (the clock read twice) + the
+# settle wait stay within 90 s.
+OOMCHECK_PREFIX = "felhom-oomcheck-"
+OOMCHECK_SCRIPT = "dd if=/dev/zero of=/dev/null bs=200M count=1"
+OOMCHECK_TIMEOUTS = {"image": 10, "clock": 5, "run": 30, "inspect": 10, "events": 10, "rm": 15}
+# Measured on demo-hp 9201 (Docker 29.8.2, 2026-10-06, audits/readback-2026-10-07/F/F1, F2): an `--until` taken right
+# after the run MISSED the oom event although OOMKilled=true; after a 2 s wait and `--until` = guest epoch + 1 it is
+# seen. Pinned by test_events_window_ends_after_the_settle_wait.
+OOMCHECK_SETTLE = 2
 # Kernel, boot and firmware packages are the SLOW lane on the host whatever their origin (`11` C3, §5.2): a host
 # reboot is needed for them to take effect, and a bad one can stop the box from booting.
 HOST_SLOW_RE = re.compile(r"^(linux-(image|headers|kbuild|modules|base)|proxmox-kernel|proxmox-default-kernel|pve-kernel|"
@@ -106,6 +127,13 @@ CRASH_GUARD_STATE = "/var/lib/felhom-crash-guard/state.json"
 # policy: replace → always written when it differs; if-absent → only when the box has none (a setting the operator may
 #         have tuned); oob → only on a box with the OOB belt (/etc/felhom-sshd exists — the installer's --no-oob leaves none).
 # Order matters: the sudoers files come LAST, so a referenced wrapper is in place before the line that allows it.
+# R-861 (agent v0.146.0): the signed agent update, verified here (mode agent_update), then the A/B flip.
+SELFUPDATE_OP = "agent_update"
+SELFUPDATE_DIR = "/var/lib/felhom-agent/selfupdate"
+SELFUPDATE_WRAPPER = "/usr/local/sbin/felhom-selfupdate-guarded"
+# The verified bytes are written HERE (a root-owned directory the agent cannot write) and only this copy reaches the A/B wrapper — never the agent's file.
+SELFUPDATE_ROOT_DIR = "/var/lib/felhom-os-apply/agent-update"
+SELFUPDATE_MAX_BYTES = 256 * 1024 * 1024
 BUNDLE_FORMAT = 1
 BUNDLE_OP = "agent_config_update"
 BUNDLE_RECORD = "/etc/felhom/config-bundle.json"     # what the box runs (0644 root; the non-root agent reports it)
@@ -121,6 +149,12 @@ BUNDLE_FILES = [
     ("/usr/local/sbin/felhom-backup-target-apply", "felhom-backup-target-apply", 0o755, "bash", "replace"),
     ("/usr/local/sbin/felhom-os-apply", "felhom-os-apply", 0o755, "python", "replace"),
     ("/usr/local/sbin/felhom-crash-guard", "felhom-crash-guard", 0o755, "python", "replace"),
+    # R-861 (agent v0.146.0): the content checker for every agent-staged file a root program reads, and the two FIXED
+    # files the agent used to install itself from /tmp (the guest pre-start hook and the shared drive parent).
+    ("/usr/local/sbin/felhom-priv-apply", "felhom-priv-apply", 0o755, "python", "replace"),
+    ("/var/lib/vz/snippets/felhom-guest-hook.sh", "felhom-guest-hook.sh", 0o755, "sh", "replace"),
+    ("/usr/local/sbin/felhom-shared-parent.sh", "felhom-shared-parent.sh", 0o755, "sh", "replace"),
+    ("/etc/systemd/system/felhom-shared-parent.service", "felhom-shared-parent.service", 0o644, "unit", "replace"),
     ("/etc/systemd/system/felhom-crash-guard.service", "felhom-crash-guard.service", 0o644, "unit", "replace"),
     ("/etc/systemd/system/felhom-crash-guard-check.service", "felhom-crash-guard-check.service", 0o644, "unit", "replace"),
     ("/etc/systemd/system/felhom-crash-guard-check.timer", "felhom-crash-guard-check.timer", 0o644, "unit", "replace"),
@@ -263,12 +297,43 @@ class Runner:
         os.replace(tmp, NONCE_FILE)
 
     def log(self, line):
-        print(line, file=sys.stderr, flush=True)
+        # R-868 (v0.144.1): the agent that reads stderr may be GONE (killed mid-pass, measured live on demo-hp
+        # 2026-10-05): the write then raises BrokenPipeError, and v0.144.0 died right there — after apt had installed
+        # everything, before the report copy was saved. A dead reader must never stop the pass; the journal still
+        # gets every line. Pinned by AgentDiesMidPass.
+        try:
+            print(line, file=sys.stderr, flush=True)
+        except OSError:
+            pass
         try:
             subprocess.run(["logger", "-t", "felhom-os-apply", line], timeout=10)
         except Exception:
             pass
 
+    def read_staged_once(self, path, owner_uid, limit):
+        """R-861: read a file the AGENT staged ONCE, as root, safely: O_NOFOLLOW (the last component may not be a
+        symlink), fstat on the opened fd (a regular file owned by owner_uid), at most `limit` bytes. The caller hashes
+        and uses exactly these bytes — never the path again (the agent owns the directory and could swap the file)."""
+        fd = os.open(path, os.O_RDONLY | os.O_NOFOLLOW | os.O_CLOEXEC)
+        try:
+            st = os.fstat(fd)
+            if not stat.S_ISREG(st.st_mode) or st.st_uid != owner_uid:
+                raise Refused("R19", f"{path} is not a regular file owned by {AGENT_USER}")
+            if st.st_size > limit:
+                raise Refused("R19", f"{path} is larger than {limit} bytes")
+            chunks, n = [], 0
+            while True:
+                b = os.read(fd, 1 << 20)
+                if not b:
+                    break
+                chunks.append(b)
+                n += len(b)
+                if n > limit:
+                    raise Refused("R19", f"{path} grew past {limit} bytes while it was read")
+            return b"".join(chunks)
+        finally:
+            os.close(fd)
+
     # ---------- host files, for the config bundle (R-840). Tests replace these with an in-memory tree. ----------
     def read_bytes(self, path):
         with open(path, "rb") as f:
@@ -358,8 +423,12 @@ class Apply:
 
     def check_plan(self, plan):
         mode = plan.get("mode", "apply")
-        if mode not in ("apply", "inventory", "health", "facts", "live-restore-on", "bundle"):
+        if mode not in ("apply", "inventory", "health", "facts", "live-restore-on", "bundle", "agent_update", "oom-check"):
             raise Refused("R11", f"unknown mode {mode!r}")
+        if mode == "agent_update":
+            if plan.get("layer") != "host":
+                raise Refused("R11", "agent_update is a host-layer mode")
+            return mode, "host", 0, "agent_update"
         if mode == "bundle":
             if plan.get("layer") != "host":
                 raise Refused("R11", "bundle is a host-layer mode")
@@ -376,6 +445,8 @@ class Apply:
             raise Refused("R11", "facts is a host-layer mode (it reads the host and the guest)")
         if mode == "live-restore-on" and layer != "guest":
             raise Refused("R11", "live-restore-on is a guest-layer mode")
+        if mode == "oom-check" and layer != "docker":
+            raise Refused("R11", "oom-check is a docker-layer mode (it checks the guest's Docker engine)")
         if plan.get("undo") and layer != "docker":
             raise Refused("R5", "an undo (downgrade) exists only for the Docker layer, inside a signed job")
         vmid = plan.get("vmid")
@@ -507,6 +578,50 @@ class Apply:
         self.burn_nonce(nonce, exp)
         return op.get("params") or {}
 
+    def agent_update(self, plan):
+        """R-861 (agent v0.146.0): the agent binary is replaced ONLY by an operator-signed agent_update, checked HERE as
+        root — signature against the root-owned signers file, op, this host, the time window, the nonce — and only the
+        staged file whose sha256 the SIGNED params pin, at the one staging path. Before v0.146.0 the agent handed the
+        sha to `felhom-selfupdate-guarded apply` itself, so a compromised agent could install any binary — and the
+        binary is what FELHOM_ESCROW and the guest hook run as root. The nonce is burned only after the flip."""
+        trust = self.load_trust()
+        params, nonce, exp = self.verify_signed(plan.get("signed"), trust, op_name=SELFUPDATE_OP, burn=False)
+        ver, sha = params.get("version"), params.get("sha256")
+        if not isinstance(ver, str) or not re.match(r"^[0-9]+\.[0-9]+\.[0-9]+$", ver):
+            raise Refused("R19", f"the signed version {ver!r} is not bare semver")
+        if not isinstance(sha, str) or not re.match(r"^[0-9a-f]{64}$", sha):
+            raise Refused("R19", "the signed sha256 is not 64 lowercase hex")
+        staged = os.path.join(SELFUPDATE_DIR, "felhom-agent-" + ver)
+        if plan.get("staged") != staged:
+            raise Refused("R19", f"the staged binary must be {staged}, got {plan.get('staged')!r}")
+        # ONE read, then never the agent's path again: hash exactly these bytes and hand the A/B wrapper a ROOT-OWNED
+        # copy of them. Hashing the agent's file and then letting the wrapper copy it by path was a race — the agent owns
+        # that directory and could swap the file between the check and the copy (found by review 2026-10-05).
+        try:
+            data = self.r.read_staged_once(staged, self.r.agent_uid(), SELFUPDATE_MAX_BYTES)
+        except OSError as e:
+            raise Refused("R19", f"cannot read the staged binary: {e}")
+        got = sha256_hex(data)
+        if got != sha:
+            raise Refused("R19", f"the staged binary's sha256 {got[:16]}… is not the signed {sha[:16]}…")
+        root_copy = os.path.join(SELFUPDATE_ROOT_DIR, "felhom-agent-" + ver)
+        self.r.put_file(root_copy, data, 0o755)
+        self.r.log(f"os-apply: AGENT-UPDATE signed by the operator: version={ver} sha={sha[:16]} — handing the root copy to the A/B wrapper")
+        try:
+            rc, out, err = self.r.host([SELFUPDATE_WRAPPER, "apply", root_copy, sha], 120)
+        finally:
+            try:
+                self.r.remove(root_copy)
+            except OSError:
+                pass
+        self.report["agent_update"] = {"version": ver, "sha256": sha, "wrapper_rc": rc,
+                                       "wrapper": (out + err).strip()[-300:]}
+        if rc != 0:
+            self.report["failed"] = {"rc": rc, "step": "agent_update", "reason": (out + err).strip()[-300:]}
+            return 3
+        self.burn_nonce(nonce, exp)
+        return 0
+
     def burn_nonce(self, nonce, exp):
         seen = self.r.read_nonces()
         if nonce in seen:
@@ -848,12 +963,18 @@ class Apply:
             return self.facts()
         if self.mode == "bundle":
             return Bundle(self).from_plan(plan)
+        if self.mode == "agent_update":
+            return self.agent_update(plan)
         if self.layer == "host":
             self.check_appliance()
         self.check_guest(self.vmid)
         log = self.r.log
         if self.mode == "live-restore-on":
             return self.live_restore_on()
+        if self.mode == "oom-check":
+            # R-528: the check alone — no apt, no engine change; check_guest above still applies.
+            self.report["oom_check"] = self.oom_check()
+            return 0
         self.who, self.allow_downgrade = ("fast", False)
         if self.layer == "docker" and self.mode == "apply":
             self.who, self.allow_downgrade = self.docker_authority(plan)
@@ -893,22 +1014,115 @@ class Apply:
             self.report["docker_engine"] = out_v.strip() if rc_v == 0 and out_v.strip() else "unknown"
         self.report["reboot_scanned"] = "reboot_needed" in self.report
         self.report["health_after"] = self.health()
+        if self.layer == "docker" and self.mode == "apply":
+            # R-528 (`09` decision 157): does the engine report a memory kill? Reported only — never the outcome.
+            self.report["oom_check"] = self.oom_check()
         return 0
 
-    def repair(self):
-        rc, before, _ = self.x(["dpkg", "--audit"])
+    # ---------- R-528: the memory-kill check ----------
+    def oom_check(self):
+        """Run one throwaway container over its memory cap in the guest and read what the engine says about it.
+        Never raises: any failure becomes result "error". The container is ALWAYS removed (finally), and a failed
+        removal is named in the detail."""
+        t = OOMCHECK_TIMEOUTS
+        res = {"result": "error", "oom_killed": False, "oom_event": False, "exit_code": None, "image": None, "detail": ""}
+        log = self.r.log
+        try:
+            rc, out, err = self.g(["docker", "inspect", "-f", "{{.Config.Image}}", "felhom-controller"], timeout=t["image"])
+        except Exception as e:
+            rc, out, err = -1, "", str(e)
+        img = out.strip() if rc == 0 else ""
+        if not img or any(c.isspace() for c in img):
+            res["detail"] = f"the controller's image could not be read (rc={rc}): {(err or out).strip()[:200]}"
+            log(f"os-apply: OOM-CHECK error — {res['detail']}")
+            return res
+        res["image"] = img
+        name = f"{OOMCHECK_PREFIX}{os.getpid()}-{os.urandom(4).hex()}"
+        notes = []
+        try:
+            t0 = self.guest_epoch()
+            if t0 is None:
+                raise RuntimeError("the guest clock could not be read")
+            rrc, rout, rerr = self.g(["docker", "run", "--name", name, "--pull", "never", "--network", "none",
+                                      "--memory", "64m", "--memory-swap", "64m", "--label", "felhom.oomcheck=1",
+                                      "--entrypoint", "sh", img, "-c", OOMCHECK_SCRIPT], timeout=t["run"])
+            self.r.sleep(OOMCHECK_SETTLE)  # the engine publishes the oom event a moment after the run returns (F1/F2)
+            t1 = self.guest_epoch()
+            if t1 is None:
+                t1 = t0 + t["run"] + OOMCHECK_SETTLE + 1
+            irc, iout, ierr = self.g(["docker", "inspect", "-f", "{{.State.OOMKilled}} {{.State.ExitCode}}", name], timeout=t["inspect"])
+            if irc != 0:
+                raise RuntimeError(f"the check container could not be inspected (run rc={rrc}: {(rerr or rout).strip()[:120]}; "
+                                   f"inspect rc={irc}: {(ierr or iout).strip()[:120]})")
+            f = iout.split()
+            res["oom_killed"] = bool(f) and f[0] == "true"
+            try:
+                res["exit_code"] = int(f[1]) if len(f) > 1 else None
+            except ValueError:
+                res["exit_code"] = None
+            erc, eout, eerr = self.g(["docker", "events", "--since", str(t0 - 1), "--until", str(t1 + 1),
+                                      "--filter", f"container={name}", "--filter", "event=oom",
+                                      "--format", "{{.Action}}"], timeout=t["events"])
+            if erc != 0:
+                notes.append(f"the event read failed (rc={erc}): {(eerr or eout).strip()[:120]}")
+            res["oom_event"] = erc == 0 and any(l.strip() == "oom" for l in eout.splitlines())
+            if res["oom_killed"] and res["oom_event"]:
+                res["result"] = "pass"
+                notes.insert(0, "the engine reported the memory kill: OOMKilled=true and the oom event")
+            else:
+                res["result"] = "fail"
+                miss = [w for w, ok in (("OOMKilled=true", res["oom_killed"]), ("the oom event", res["oom_event"])) if not ok]
+                notes.insert(0, f"the engine did not report the memory kill: missing {' and '.join(miss)} (exit code {res['exit_code']})")
+        except Exception as e:
+            res["result"] = "error"
+            notes.insert(0, f"the check could not finish: {type(e).__name__}: {str(e)[:200]}")
+        finally:
+            try:
+                mrc, mout, merr = self.g(["docker", "rm", "-f", name], timeout=t["rm"])
+                if mrc != 0 and "no such container" not in (merr + mout).lower():
+                    notes.append(f"the check container {name} could not be removed (rc={mrc}): {(merr or mout).strip()[:120]}")
+            except Exception as e:
+                notes.append(f"the check container {name} could not be removed: {type(e).__name__}: {str(e)[:120]}")
+        res["detail"] = "; ".join(notes)
+        log(f"os-apply: OOM-CHECK result={res['result']} oom_killed={res['oom_killed']} oom_event={res['oom_event']} "
+            f"exit={res['exit_code']} image={img} — {res['detail']}")
+        return res
+
+    def guest_epoch(self):
+        rc, out, _ = self.g(["date", "+%s"], timeout=OOMCHECK_TIMEOUTS["clock"])
+        try:
+            return int(out.strip()) if rc == 0 else None
+        except ValueError:
+            return None
+
+    def dpkg_state(self):
+        """`dpkg --audit` AND dpkg's update journal, in ONE call (R-876, agent v0.145.0). A crash in the middle of an
+        install can leave `/var/lib/dpkg/updates/` non-empty while `--audit` reads clean — measured on demo-hp
+        2026-10-05 — and that journal is exactly what apt refuses on ("dpkg was interrupted"). One `sh -c` with a
+        constant script keeps R-845's speed: a clean pass still costs one call here, as before."""
+        rc, out, _ = self.x(["sh", "-c", DPKG_STATE_SCRIPT])
+        audit, _, journal = out.partition(JOURNAL_MARK + "\n")
+        return audit, [l for l in journal.split() if l]
+
+    def repair(self, force=False):
+        before, journal = self.dpkg_state()
         configured = len([l for l in before.splitlines() if l.startswith(" ")])
         fixed = 0
-        after = ""
-        if before.strip():  # nothing half-done → nothing to run (R-845: two calls saved on every clean pass)
+        after, journal_after = "", []
+        # nothing half-done and no update journal → nothing to run (R-845: two calls saved on every clean pass).
+        # R-876: the JOURNAL counts too, and `force` (apt said "dpkg was interrupted") always repairs.
+        if before.strip() or journal or force:
             self.x(APT_ENV + ["dpkg", "--configure", "-a", "--force-confold"])
             rc2, out, err = self.x(APT_ENV + ["apt-get", "-f", "install", "-y", "-q"] + DPKG_OPTS)
-            _, after, _ = self.x(["dpkg", "--audit"])
+            after, journal_after = self.dpkg_state()
             fixed = len(re.findall(r"^Setting up ", out, re.M))
-        self.report["repair"] = {"half_configured_before": configured, "fixed": fixed, "clean_after": after.strip() == ""}
-        self.r.log(f"os-apply: REPAIR configured={configured} fixed={fixed}")
+        self.report["repair"] = {"half_configured_before": configured, "journal_before": len(journal), "fixed": fixed,
+                                 "clean_after": after.strip() == "" and not journal_after}
+        self.r.log(f"os-apply: REPAIR configured={configured} journal={len(journal)} fixed={fixed}" + (" forced" if force else ""))
         if after.strip():
             raise Refused("R13", "dpkg is still broken after the repair: " + after.strip().splitlines()[0])
+        if journal_after:
+            raise Refused("R13", f"dpkg's update journal is still not empty after the repair ({len(journal_after)} file(s))")
 
     def pending_fast(self):
         """Ring 0 (select pending-fast): every pending upgrade of an INSTALLED package whose every origin is Debian /
@@ -1025,6 +1239,12 @@ class Apply:
                 raise Refused("R8", f"free space {free} B is below max(500 MB, 3 x download {need} B)")
             t0 = time.time()
             rc, out, err = self.x(APT_ENV + ["apt-get", "-y", "-q"] + DPKG_OPTS + args)
+            if rc != 0 and "dpkg was interrupted" in (out + err):
+                # R-876 (belt): apt says dpkg was interrupted although the repair found nothing — repair and
+                # retry ONCE. Never a loop.
+                self.r.log("os-apply: INTERRUPTED apt says dpkg was interrupted — repairing and retrying once")
+                self.repair(force=True)
+                rc, out, err = self.x(APT_ENV + ["apt-get", "-y", "-q"] + DPKG_OPTS + args)
             secs = time.time() - t0
             # dpkg says "Installing new version of config file X" when X was NOT changed locally (the package's new
             # version is taken), and "Configuration file 'X'" + "Keeping old config file" when it was (--force-confold
@@ -1335,6 +1555,11 @@ class Bundle:
         if rc != 2 or "usage" not in (out + err):
             raise Refused("R18", f"the installed felhom-selfupdate-guarded does not answer with its usage (rc={rc})")
         sc["selfupdate"] = "usage ok"
+        if "/usr/local/sbin/felhom-priv-apply" in {p[0] for p in plan if p[3] != "skipped"}:
+            rc, out, err = self.r.host(["/usr/bin/python3", "/usr/local/sbin/felhom-priv-apply", "--self-check"], 60)
+            if rc != 0 or "felhom-priv-apply ok" not in out:
+                raise Refused("R18", f"the installed felhom-priv-apply does not answer its self-check (rc={rc})")
+            sc["priv_apply"] = out.strip()[:80]
         dests = {p[0] for p in plan if p[3] != "skipped"}
         if "/usr/local/sbin/felhom-crash-guard" in dests:
             rc, out, err = self.r.host(["/usr/local/sbin/felhom-crash-guard", "status"], 60)
@@ -1453,7 +1678,11 @@ def main(argv, runner=None, environ=None):
             r.save_report(argv[2], a.report)
         except Exception as e:
             r.log(f"os-apply: the report copy could not be saved (the run is unaffected): {e}")
-    print("OSAPPLY-REPORT " + json.dumps(a.report, sort_keys=True))
+    try:
+        print("OSAPPLY-REPORT " + json.dumps(a.report, sort_keys=True), flush=True)
+    except OSError:
+        # R-868: nobody reads stdout any more (the agent was killed); the copy above carries the report.
+        sys.stdout = open(os.devnull, "w")
     return rc
 
 

configs/felhom-priv-apply

@@ -0,0 +1,419 @@
+#!/usr/bin/python3
+"""felhom-priv-apply — the ROOT half of every file the agent writes into a root-read place (R-861, `03` §3.1).
+
+Install as /usr/local/sbin/felhom-priv-apply (0755 root:root) — it rides the signed config bundle (R-840).
+
+WHY IT EXISTS. Until agent v0.146.0 the agent's sudoers let it `install` a file it had written itself into a place a
+root program reads: a systemd .mount unit (a bind mount of an agent-owned directory over /etc/sudoers.d is a root
+shell), a dnsmasq drop-in (`dhcp-script=` runs as root), the WireGuard config (`PostUp=` runs as root) and the OOB
+sshd config (`AuthorizedKeysFile` + `StrictModes no`). A compromised agent PROCESS was therefore root on its host.
+Now the agent stages the file and this wrapper — root-owned, delivered only by an operator-signed bundle — checks
+the CONTENT against the exact grammar the agent's own renderers produce, and refuses anything else. The agent can no
+longer name the destination: each verb has a fixed source and a fixed (or strictly named) destination.
+
+Verbs (each one sudoers line, exact-match pattern):
+  unit <name>              /var/lib/felhom-agent/units/<name> -> /etc/systemd/system/<name>   (.mount | .automount)
+  dnsmasq <tmp> <name>     /tmp/felhom-resolver-<digits>.conf -> /etc/dnsmasq.d/felhom-<...>.conf
+  wg                       /var/lib/felhom-agent/wg/wg-felhom.conf -> /etc/wireguard/wg-felhom.conf (0600)
+  sshd-config              /var/lib/felhom-agent/felhom-sshd/sshd_config -> /etc/felhom-sshd/sshd_config
+  sshd-key                 /var/lib/felhom-agent/felhom-sshd/authorized_keys.felhom-op -> /etc/felhom-sshd/authorized_keys/felhom-op
+  --self-check             prints "felhom-priv-apply ok verbs=..." (the bundle's self-check)
+
+Exit codes: 0 installed (or already identical), 2 usage, 3 refused (content or source), 4 install failed.
+Every refusal is logged to the journal (tag felhom-priv-apply) with its rule; file CONTENT is never logged.
+Pinned by configs/test_felhom_priv_apply.py (one test per rule, red-proofs in the R-861 audit).
+"""
+import ipaddress
+import os
+import re
+import stat
+import subprocess
+import sys
+
+AGENT_USER = "felhom-agent"
+STATE = "/var/lib/felhom-agent"
+UNITS_SRC = STATE + "/units"
+UNIT_DIR = "/etc/systemd/system"
+DNSMASQ_DIR = "/etc/dnsmasq.d"
+WG_SRC, WG_DEST = STATE + "/wg/wg-felhom.conf", "/etc/wireguard/wg-felhom.conf"
+SSHD_SRC, SSHD_DEST = STATE + "/felhom-sshd/sshd_config", "/etc/felhom-sshd/sshd_config"
+KEY_SRC, KEY_DEST = STATE + "/felhom-sshd/authorized_keys.felhom-op", "/etc/felhom-sshd/authorized_keys/felhom-op"
+MAX_BYTES = 64 * 1024
+VERBS = ("unit", "dnsmasq", "wg", "sshd-config", "sshd-key")
+
+UNIT_NAME_RE = re.compile(r"^mnt-[A-Za-z0-9_.\\-]+\.(mount|automount)$")
+DNSMASQ_TMP_RE = re.compile(r"^/tmp/felhom-resolver-[0-9]+\.conf$")
+DNSMASQ_NAME_RE = re.compile(r"^felhom-[a-z0-9][a-z0-9._-]*\.conf$")
+SEG = r"[A-Za-z0-9_-][A-Za-z0-9_.-]*"
+WHERE_RE = re.compile(r"^/mnt/(felhom-drives/)?" + SEG + r"$")
+UUID_RE = re.compile(r"^[A-Fa-f0-9]{4,}(-[A-Fa-f0-9]+){0,4}$")
+HOST_RE = re.compile(r"^[A-Za-z0-9._:-]{1,255}$")
+NET_PATH_RE = re.compile(r"^[A-Za-z0-9._/@+-]{1,512}$")
+OPT_RE = re.compile(r"^[A-Za-z0-9_.:/@+-]+(=[A-Za-z0-9_.:/@+-]+)?$")
+LOCAL_TYPES = {"ext4", "xfs", "btrfs", "exfat", "vfat", "ntfs3", "ntfs"}
+NET_TYPES = {"nfs", "nfs4", "cifs"}
+# Options that turn a device mount into something else, or let set-uid/device files act on the host.
+FORBIDDEN_OPTS = {"bind", "rbind", "move", "rmove", "remount", "suid", "dev", "user", "users", "owner", "group",
+                  "x-mount.mkdir", "helper"}
+DESC_RE = re.compile(r"^[^\x00-\x1f\x7f]{0,200}$")
+WG_KEY_RE = re.compile(r"^[A-Za-z0-9+/]{42}[AEIMQUYcgkosw480]=$")
+KEY_LINE_RE = re.compile(r"^(ssh-ed25519|ssh-rsa|ecdsa-sha2-nistp(256|384|521)|sk-ssh-ed25519@openssh\.com) "
+                         r"[A-Za-z0-9+/]+={0,3}( [ -~]{0,200})?$")
+
+
+class Refused(Exception):
+    def __init__(self, rule, reason):
+        super().__init__(reason)
+        self.rule, self.reason = rule, reason
+
+
+class Host:
+    """Every filesystem / process effect, so the tests can play the box in memory."""
+
+    def agent_uid(self):
+        import pwd
+        return pwd.getpwnam(AGENT_USER).pw_uid
+
+    def read_source(self, path):
+        """The staged file: a REGULAR file owned by the agent, never a symlink, at most MAX_BYTES."""
+        try:
+            fd = os.open(path, os.O_RDONLY | os.O_NOFOLLOW | os.O_CLOEXEC)
+        except OSError as e:
+            raise Refused("P1", f"cannot open the staged file {path}: {e.strerror}")
+        try:
+            st = os.fstat(fd)
+            if not stat.S_ISREG(st.st_mode):
+                raise Refused("P1", f"{path} is not a regular file")
+            if st.st_uid != self.agent_uid():
+                raise Refused("P1", f"{path} is not owned by {AGENT_USER}")
+            if st.st_size > MAX_BYTES:
+                raise Refused("P1", f"{path} is larger than {MAX_BYTES} bytes")
+            with os.fdopen(fd, "rb") as f:
+                fd = -1
+                return f.read(MAX_BYTES + 1)
+        finally:
+            if fd >= 0:
+                os.close(fd)
+
+    def read_dest(self, path):
+        try:
+            with open(path, "rb") as f:
+                return f.read()
+        except OSError:
+            return None
+
+    def install(self, dest, data, mode):
+        """Atomic, root-owned: a temp file beside the destination, fsync, rename."""
+        d = os.path.dirname(dest)
+        os.makedirs(d, mode=0o755, exist_ok=True)
+        tmp = os.path.join(d, f".{os.path.basename(dest)}.felhom-new.{os.getpid()}")
+        fd = os.open(tmp, os.O_WRONLY | os.O_CREAT | os.O_EXCL | os.O_NOFOLLOW, 0o600)
+        try:
+            with os.fdopen(fd, "wb") as f:
+                f.write(data)
+                f.flush()
+                os.fchown(f.fileno(), 0, 0)
+                os.fchmod(f.fileno(), mode)
+                os.fsync(f.fileno())
+            os.replace(tmp, dest)
+        except BaseException:
+            try:
+                os.remove(tmp)
+            except OSError:
+                pass
+            raise
+
+    def log(self, line):
+        print(line, file=sys.stderr)
+        try:
+            subprocess.run(["logger", "-t", "felhom-priv-apply", "--", line], timeout=10, check=False)
+        except (OSError, subprocess.SubprocessError):
+            pass
+
+
+def systemd_escape_path(path):
+    """`systemd-escape --path`: strip the slashes at both ends, `/` -> `-`, every byte outside [A-Za-z0-9:_.] (and a
+    leading `.`) -> `\\xNN`."""
+    p = path.strip("/")
+    out = []
+    for i, ch in enumerate(p):
+        if ch == "/":
+            out.append("-")
+        elif (ch.isascii() and (ch.isalnum() or ch in ":_.")) and not (i == 0 and ch == "."):
+            out.append(ch)
+        else:
+            out.extend("\\x%02x" % b for b in ch.encode())
+    return "".join(out)
+
+
+def text_of(data, what):
+    if len(data) > MAX_BYTES:
+        raise Refused("P1", f"{what} is too large")
+    try:
+        text = data.decode("utf-8")
+    except UnicodeDecodeError:
+        raise Refused("P2", f"{what} is not UTF-8 text")
+    if "\x00" in text or "\r" in text:
+        raise Refused("P2", f"{what} carries a NUL or CR byte")
+    return text
+
+
+def parse_ini(text, what):
+    """[Section] / Key=Value / comments / blank lines. A key outside a section or a repeated key is refused."""
+    sections, cur = {}, None
+    for n, raw in enumerate(text.split("\n"), 1):
+        line = raw.strip()
+        if line.endswith("\\"):
+            # systemd joins a line ending in a backslash with the next one; this parser does not. Refused, so the two
+            # can never read the same bytes differently (review 2026-10-05).
+            raise Refused("U2", f"{what}: line {n} ends with a backslash (a continuation)")
+        if not line or line.startswith("#") or line.startswith(";"):
+            continue
+        m = re.match(r"^\[([A-Za-z]+)\]$", line)
+        if m:
+            cur = m.group(1)
+            if cur in sections:
+                raise Refused("U2", f"{what}: section [{cur}] twice")
+            sections[cur] = {}
+            continue
+        if cur is None or "=" not in line:
+            raise Refused("U2", f"{what}: line {n} is not Key=Value inside a section")
+        k, v = line.split("=", 1)
+        k, v = k.strip(), v.strip()
+        if k in sections[cur]:
+            raise Refused("U2", f"{what}: {cur}.{k} given twice")
+        sections[cur][k] = v
+    return sections
+
+
+# ---------- the unit verb ----------
+def check_unit(name, text):
+    if not UNIT_NAME_RE.match(name) or "/" in name:
+        raise Refused("U1", f"unit name {name!r} is not mnt-<escaped path>.mount|.automount")
+    kind = "automount" if name.endswith(".automount") else "mount"
+    s = parse_ini(text, name)
+    body = "Automount" if kind == "automount" else "Mount"
+    # [Unit] holds ONLY what the renderers write: Description, and After=local-fs-pre.target on a local mount. A
+    # Wants=/Requires=/Before= naming any unit would start it with the mount (Wants=reboot.target — found by review
+    # 2026-10-05), so none of them is accepted.
+    allowed = {"Unit": {"Description", "After"},
+               body: {"Where", "TimeoutIdleSec"} if kind == "automount" else {"What", "Where", "Type", "Options"},
+               "Install": {"WantedBy"}}
+    for sec, keys in s.items():
+        if sec not in allowed:
+            raise Refused("U2", f"{name}: section [{sec}] is not allowed")
+        bad = set(keys) - allowed[sec]
+        if bad:
+            raise Refused("U2", f"{name}: [{sec}] key(s) {sorted(bad)} not allowed")
+    u = s.get("Unit", {})
+    if not DESC_RE.match(u.get("Description", "")):
+        raise Refused("U2", f"{name}: Description has control characters")
+    if "After" in u and u["After"] != "local-fs-pre.target":
+        raise Refused("U2", f"{name}: After= may only be local-fs-pre.target")
+    inst = s.get("Install", {})
+    if inst and inst.get("WantedBy") != "multi-user.target":
+        raise Refused("U2", f"{name}: WantedBy must be multi-user.target")
+    m = s.get(body)
+    if not m or "Where" not in m:
+        raise Refused("U3", f"{name}: no [{body}] Where=")
+    where = m["Where"]
+    if not WHERE_RE.match(where):
+        raise Refused("U3", f"{name}: Where={where} is not /mnt/<name> or /mnt/felhom-drives/<name>")
+    if systemd_escape_path(where) + "." + kind != name:
+        raise Refused("U3", f"{name}: the unit name does not match Where={where}")
+    if kind == "automount":
+        t = m.get("TimeoutIdleSec", "")
+        if t and not re.match(r"^[0-9]{1,6}$", t):
+            raise Refused("U2", f"{name}: TimeoutIdleSec must be seconds")
+        return
+    what, typ = m.get("What", ""), m.get("Type", "")
+    opts = [o for o in m.get("Options", "").split(",") if o]
+    net = False
+    mu = re.match(r"^/dev/disk/by-uuid/(.+)$", what)
+    if mu:
+        if not UUID_RE.match(mu.group(1)):
+            raise Refused("U4", f"{name}: What= is not a filesystem UUID")
+        if typ and typ not in LOCAL_TYPES:
+            raise Refused("U4", f"{name}: Type={typ} is not a local filesystem")
+    else:
+        net = True
+        if typ not in NET_TYPES:
+            raise Refused("U4", f"{name}: What= is neither /dev/disk/by-uuid/<uuid> nor a network source with Type=nfs/nfs4/cifs")
+        if typ == "cifs":
+            mm = re.match(r"^//([^/]+)/(.+)$", what)
+        else:
+            mm = re.match(r"^([^/:][^:]*):(/.*)$", what)
+        if not mm or not HOST_RE.match(mm.group(1)) or not NET_PATH_RE.match(mm.group(2)) or ".." in mm.group(2).split("/"):
+            raise Refused("U4", f"{name}: What= is not a clean {typ} source")
+        if not where.startswith("/mnt/felhom-drives/"):
+            raise Refused("U3", f"{name}: a network share mounts only under /mnt/felhom-drives/")
+    for o in opts:
+        if not OPT_RE.match(o):
+            raise Refused("U5", f"{name}: mount option {o!r} has characters a mount option never needs")
+        if o.split("=", 1)[0].lower() in FORBIDDEN_OPTS or o.lower().startswith("x-mount."):
+            raise Refused("U5", f"{name}: mount option {o.split('=', 1)[0]!r} is not allowed")
+    if net and not {"nosuid", "nodev"} <= set(opts):
+        # A network server is outside the box: a set-uid file on it must never run as root here.
+        raise Refused("U5", f"{name}: a network share must carry nosuid,nodev")
+
+
+# ---------- dnsmasq ----------
+def _ip(v, v6=True):
+    try:
+        a = ipaddress.ip_address(v)
+    except ValueError:
+        return False
+    return v6 or a.version == 4
+
+
+DOMAIN_RE = re.compile(r"^[A-Za-z0-9]([A-Za-z0-9-]{0,62})(\.[A-Za-z0-9]([A-Za-z0-9-]{0,62}))*$")
+
+
+def check_dnsmasq(text):
+    for n, raw in enumerate(text.split("\n"), 1):
+        line = raw.strip()
+        if not line or line.startswith("#"):
+            continue
+        if line in ("bind-interfaces", "no-resolv"):
+            continue
+        k, _, v = line.partition("=")
+        if k == "listen-address" and _ip(v, v6=False):
+            continue
+        if k == "server" and (_ip(v) or (v.count("#") == 1 and _ip(v.split("#")[0]) and v.split("#")[1].isdigit())):
+            continue
+        m = re.match(r"^/([^/]+)/$", v)
+        if k == "local" and m and DOMAIN_RE.match(m.group(1)):
+            continue
+        m = re.match(r"^/([^/]+)/([^/]+)$", v)
+        if k == "address" and m and DOMAIN_RE.match(m.group(1)) and _ip(m.group(2), v6=False):
+            continue
+        raise Refused("D1", f"dnsmasq line {n} ({k or line[:20]!r}) is not one the resolver writes")
+
+
+# ---------- WireGuard ----------
+def check_wg(text):
+    s = parse_ini(text, "wg-felhom.conf")
+    if set(s) != {"Interface", "Peer"}:
+        raise Refused("W1", "wg-felhom.conf must hold exactly [Interface] and [Peer]")
+    i, p = s["Interface"], s["Peer"]
+    if set(i) - {"PrivateKey", "Address", "MTU"} or set(p) - {"PublicKey", "Endpoint", "AllowedIPs", "PersistentKeepalive"}:
+        raise Refused("W1", "wg-felhom.conf carries a key the agent never writes (PostUp/PreUp/... run as root)")
+    if not WG_KEY_RE.match(i.get("PrivateKey", "")) or not WG_KEY_RE.match(p.get("PublicKey", "")):
+        raise Refused("W2", "a WireGuard key is not 32 bytes of base64")
+    try:
+        a = ipaddress.ip_network(i.get("Address", ""), strict=False)
+        if a.version != 4 or a.prefixlen != 32:
+            raise ValueError
+        if not (1280 <= int(i.get("MTU", "1280")) <= 1500):
+            raise ValueError
+        host, _, port = p.get("Endpoint", "").rpartition(":")
+        if ipaddress.ip_address(host).version != 4 or not (1 <= int(port) <= 65535):
+            raise ValueError
+        for n in p.get("AllowedIPs", "").split(","):
+            if ipaddress.ip_network(n.strip(), strict=True).prefixlen != 32:
+                raise ValueError
+        if not (0 <= int(p.get("PersistentKeepalive", "25")) <= 3600):
+            raise ValueError
+    except ValueError:
+        raise Refused("W2", "an Address/MTU/Endpoint/AllowedIPs/PersistentKeepalive value is not what the agent renders")
+
+
+# ---------- OOB sshd ----------
+def render_sshd(port):
+    """Byte-identical to felhomsshd.renderConfig (internal/felhomsshd/config.go) — pinned by a Go test."""
+    return ("# felhom OOB sshd — agent-managed (H1); DO NOT EDIT\n"
+            f"Port {port}\n"
+            "ListenAddress 0.0.0.0\n"
+            "ListenAddress ::\n"
+            "HostKey /etc/felhom-sshd/ssh_host_ed25519_key\n"
+            "PidFile /run/felhom-sshd.pid\n"
+            "AuthorizedKeysFile /etc/felhom-sshd/authorized_keys/%u\n"
+            "PasswordAuthentication no\n"
+            "PermitRootLogin prohibit-password\n"
+            "PubkeyAuthentication yes\n"
+            "KbdInteractiveAuthentication no\n"
+            "UsePAM yes\n"
+            "AllowUsers root felhom-op\n"
+            "X11Forwarding no\n"
+            "Subsystem sftp internal-sftp\n")
+
+
+def check_sshd(text):
+    m = re.search(r"^Port ([0-9]{1,5})$", text, re.M)
+    if not m or not (1 <= int(m.group(1)) <= 65535) or int(m.group(1)) == 22:
+        raise Refused("S1", "sshd_config has no Port (or claims :22, the household's sshd)")
+    if text != render_sshd(int(m.group(1))):
+        raise Refused("S1", "sshd_config differs from the one fixed template (only the Port may vary)")
+
+
+def check_key(text):
+    lines = [l for l in text.split("\n") if l.strip()]
+    if len(lines) > 1:
+        raise Refused("S2", "felhom-op's authorized_keys holds more than one key")
+    if lines and not KEY_LINE_RE.match(lines[0]):
+        raise Refused("S2", "the key line is not a plain public key (no options such as command= or from=)")
+
+
+# ---------- main ----------
+def plan(argv):
+    """(verb, source, dest, mode, checker) for an argv, or Refused("A1")."""
+    if not argv or argv[0] not in VERBS:
+        raise Refused("A1", "usage: felhom-priv-apply unit <name> | dnsmasq <tmp> <name> | wg | sshd-config | sshd-key")
+    v, rest = argv[0], argv[1:]
+    if v == "unit" and len(rest) == 1:
+        if not UNIT_NAME_RE.match(rest[0]):
+            raise Refused("U1", f"unit name {rest[0]!r} is not mnt-<escaped path>.mount|.automount")
+        return v, os.path.join(UNITS_SRC, rest[0]), os.path.join(UNIT_DIR, rest[0]), 0o644, lambda t: check_unit(rest[0], t)
+    if v == "dnsmasq" and len(rest) == 2:
+        if not DNSMASQ_TMP_RE.match(rest[0]) or not DNSMASQ_NAME_RE.match(rest[1]):
+            raise Refused("D2", "dnsmasq wants /tmp/felhom-resolver-<digits>.conf and felhom-<name>.conf")
+        return v, rest[0], os.path.join(DNSMASQ_DIR, rest[1]), 0o644, check_dnsmasq
+    if v == "wg" and not rest:
+        return v, WG_SRC, WG_DEST, 0o600, check_wg
+    if v == "sshd-config" and not rest:
+        return v, SSHD_SRC, SSHD_DEST, 0o644, check_sshd
+    if v == "sshd-key" and not rest:
+        return v, KEY_SRC, KEY_DEST, 0o644, check_key
+    raise Refused("A1", f"wrong arguments for {v}")
+
+
+def main(argv, host=None):
+    host = host or Host()
+    if argv == ["--self-check"]:
+        print("felhom-priv-apply ok verbs=" + ",".join(VERBS))
+        return 0
+    if len(argv) >= 3 and argv[0] == "--check":
+        # CHECK ONLY (tests and the Go contract tests): `--check <verb> [<name>] <file>` validates <file> as that
+        # verb would and installs nothing. Not in sudoers. Prints OK or the rule; never the content.
+        verb, file = argv[1], argv[-1]
+        try:
+            _, _, _, _, checker = plan([verb] + argv[2:-1] if verb != "dnsmasq" else
+                                       [verb, "/tmp/felhom-resolver-1.conf", argv[2]])
+            with open(file, "rb") as f:
+                checker(text_of(f.read(), file))
+        except Refused as e:
+            print(f"REFUSED [{e.rule}] {e.reason}")
+            return 3
+        print("OK")
+        return 0
+    try:
+        verb, src, dest, mode, checker = plan(argv)
+        data = host.read_source(src)
+        checker(text_of(data, src))
+    except Refused as e:
+        host.log(f"felhom-priv-apply: REFUSED [{e.rule}] {' '.join(argv)[:160]}: {e.reason}")
+        return 2 if e.rule == "A1" else 3
+    if host.read_dest(dest) == data:
+        host.log(f"felhom-priv-apply: SAME {verb} {dest}")
+        return 0
+    try:
+        host.install(dest, data, mode)
+    except OSError as e:
+        host.log(f"felhom-priv-apply: FAILED {verb} {dest}: {e}")
+        return 4
+    host.log(f"felhom-priv-apply: INSTALLED {verb} {dest} ({len(data)} bytes)")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main(sys.argv[1:]))

configs/felhom-selfupdate-guarded

@@ -24,6 +24,11 @@ set -u
 BIN=/usr/local/bin/felhom-agent
 PREV=$BIN.prev
 STAGING=/var/lib/felhom-agent/selfupdate
+# R-861 (agent v0.146.1): `apply` takes ONLY the root-owned copy felhom-os-apply writes after it has verified the
+# operator's signature and hashed exactly those bytes (mode agent_update). The agent cannot call `apply` any more (it
+# left the sudoers), and the agent's own staging dir is no longer accepted: a file in a directory the agent owns can be
+# swapped between this script's sha check and its copy.
+ROOT_STAGING=/var/lib/felhom-os-apply/agent-update
 PENDING=$STAGING/pending.json
 UNIT=felhom-agent.service
 
@@ -42,11 +47,14 @@ apply)
         log "refusing apply: usage: apply <staged> <sha256>"
         exit 2
     fi
-    # Root-side path confinement: the staged binary MUST live in the agent's staging dir.
+    # Root-side path confinement: the staged binary MUST be felhom-os-apply's root-owned copy (R-861).
     case "$staged" in
-        "$STAGING"/*) ;;
-        *) log "refusing apply: staged path outside $STAGING: $staged"; exit 1 ;;
+        "$ROOT_STAGING"/*) ;;
+        *) log "refusing apply: staged path outside $ROOT_STAGING: $staged"; exit 1 ;;
     esac
+    if [ -L "$staged" ] || [ "$(stat -c %u "$staged" 2>/dev/null)" != "0" ]; then
+        log "refusing apply: $staged is a symlink or not root-owned"; exit 1
+    fi
     case "$staged" in
         *..*) log "refusing apply: staged path contains '..'"; exit 1 ;;
     esac

configs/felhom-shared-parent.service

@@ -0,0 +1,13 @@
+[Unit]
+Description=Felhom stable drive parent (shared bind for live drive hot-swap)
+After=local-fs.target
+Before=pve-guests.service
+ConditionPathExists=/usr/local/sbin/felhom-shared-parent.sh
+
+[Service]
+Type=oneshot
+RemainAfterExit=yes
+ExecStart=/usr/local/sbin/felhom-shared-parent.sh
+
+[Install]
+WantedBy=pve-guests.service multi-user.target

configs/felhom-shared-parent.sh

@@ -0,0 +1,15 @@
+#!/bin/sh
+# felhom stable drive parent: a SHARED bind so the agent can swap backing drives underneath it and the
+# guest sees the change live (no restart). MUST run before pve-guests so the guest's parent bind inherits
+# the shared peer group (slave). Installed + enabled by felhom-agent. Idempotent.
+set -e
+mkdir -p /mnt/felhom-drives
+# Isolate + share ONLY when first creating the self-bind (a fresh boot). The self-bind inherits the root
+# mount's shared peer group, so make-private detaches it (else binds under it DOUBLE via the root peer),
+# then make-shared gives it its own group whose only slave is the guest's parent bind. Re-running this on
+# an existing parent would churn the peer-group id and orphan the guest's slave — so guard on mountpoint.
+if ! mountpoint -q /mnt/felhom-drives; then
+  mount --bind /mnt/felhom-drives /mnt/felhom-drives
+  mount --make-private /mnt/felhom-drives
+  mount --make-shared /mnt/felhom-drives
+fi

configs/test_felhom_config_bundle.py

@@ -94,6 +94,14 @@ class Box:
             raise OSError("no such file")
         return self.files[p]
 
+    def read_staged_once(self, p, owner_uid, limit):
+        if p not in self.files:
+            raise OSError("no such file")
+        if self.uids[p] != owner_uid:
+            raise osapply.Refused("R19", f"{p} is not a regular file owned by felhom-agent")
+        self.staged_reads = getattr(self, "staged_reads", 0) + 1
+        return self.files[p]
+
     def stat(self, p):
         if p not in self.files:
             raise OSError("no such file")
@@ -129,6 +137,9 @@ class Box:
             return (1, "", "parse error") if self.visudo_fail else (0, "ok", "")
         if argv[:2] == ["sudo", "-n"]:
             return 0, self.sudo_l, ""
+        if argv[-2:] == ["/usr/local/sbin/felhom-priv-apply", "--self-check"]:
+            body = self.files.get("/usr/local/sbin/felhom-priv-apply", b"")
+            return (0, "felhom-priv-apply ok verbs=unit\n", "") if b"VERBS" in body else (1, "", "boom")
         if argv[-1] == "--self-check":
             body = self.files.get("/usr/local/sbin/felhom-os-apply", b"")
             return (0, "felhom-os-apply ok bundle-format=1 files=22\n", "") if b"BUNDLE_OP" in body else (1, "", "boom")
@@ -535,11 +546,191 @@ class Builder(unittest.TestCase):
                                r"/etc/felhom/[a-z.-]+)", text))
         agent_writes = {"/usr/local/sbin/felhom-shared-parent", "/etc/systemd/system/felhom-shared-parent.service"}
         trust = {osapply.TRUST_FILE, osapply.TRUST_SIGNERS, osapply.TRUST_SIGNERS + ".tmp", osapply.BUNDLE_RECORD}
-        missing = sorted(p for p in found if p not in osapply.BUNDLE_DESTS and p not in agent_writes | trust)
+        # Written by the appliance ISO's first boot (felhom.eu scripts/iso/felhom-bootstrap.sh), never by the installer:
+        # since installer 1.32.0 (R-275) the uninstall only NAMES them under KEPT.
+        iso_writes = {"/etc/felhom/.bootstrap-done", "/etc/felhom/appliance-pairing-code"}
+        missing = sorted(p for p in found if p not in osapply.BUNDLE_DESTS and p not in agent_writes | trust | iso_writes)
         self.assertEqual(missing, [], "the installer writes these root files, but the bundle does not carry them")
         # the limits drop-in is named through $AGENT_UNIT in the installer
         self.assertIn("/etc/systemd/system/felhom-agent.service.d/felhom-agent-limits.conf", osapply.BUNDLE_DESTS)
 
 
+
+# ---------- R-861 (agent v0.146.0): the agent binary only by an operator-signed agent_update, checked as root ----------
+STAGED = "/var/lib/felhom-agent/selfupdate/felhom-agent-0.146.0"
+NEW_BIN = b"\x7fELF the new agent"
+
+
+def update_job(sha, version="0.146.0", op="agent_update", nonce="u1", host=HOST):
+    blob = json.dumps({"expires_at": "2026-10-04T12:30:00Z", "issued_at": "2026-10-04T11:50:00Z", "key_id": "felhom-op-1",
+                       "nonce": nonce, "op": op, "params": {"sha256": sha, "version": version},
+                       "target": {"guest_id": "", "host_id": host}}, sort_keys=True).encode()
+    return {"blob_b64": base64.b64encode(blob).decode(), "sig": "-----BEGIN SSH SIGNATURE-----\nx\n-----END SSH SIGNATURE-----\n"}
+
+
+def update_box(job, staged=STAGED, content=NEW_BIN):
+    box = Box(b"{}", None)
+    box.put(PLAN, json.dumps({"release_id": "agent-0.146.0", "layer": "host", "mode": "agent_update",
+                              "signed": job, "staged": staged}).encode(), 0o600, uid=999)
+    box.put(STAGED, content, 0o755, uid=999)
+    return box
+
+
+def wrapper_calls(box):
+    return [c for c in box.calls if c and c[0] == osapply.SELFUPDATE_WRAPPER and c[1:2] == ["apply"]]
+
+
+class AgentUpdate(unittest.TestCase):
+    """RED-PROOF: make agent_update skip verify_signed → test_a_bad_signature_never_reaches_the_wrapper fails."""
+
+    def test_signed_update_flips_and_burns_the_nonce(self):
+        sha = hashlib.sha256(NEW_BIN).hexdigest()
+        box = update_box(update_job(sha))
+        rc, rep = run(box)
+        self.assertEqual(rc, 0, rep)
+        root_copy = osapply.SELFUPDATE_ROOT_DIR + "/felhom-agent-0.146.0"
+        # the wrapper gets the ROOT-OWNED copy of the bytes that were hashed — never the agent's path (review 2026-10-05)
+        self.assertEqual(wrapper_calls(box), [[osapply.SELFUPDATE_WRAPPER, "apply", root_copy, sha]])
+        self.assertIn(root_copy, box.writes)
+        self.assertNotIn(root_copy, box.files, "the root copy is removed after the flip")
+        self.assertEqual(box.staged_reads, 1, "the agent's file is read exactly once")
+        self.assertIn("u1", box.nonces)
+        self.assertEqual(rep["agent_update"]["version"], "0.146.0")
+
+    def test_a_staged_file_the_agent_does_not_own_is_refused(self):
+        sha = hashlib.sha256(NEW_BIN).hexdigest()
+        box = update_box(update_job(sha))
+        box.uids[STAGED] = 0
+        rc, rep = run(box)
+        self.assertEqual((rc, rep["refused"]["code"]), (2, "R19"), rep)
+        self.assertEqual(wrapper_calls(box), [])
+
+    def test_a_bad_signature_never_reaches_the_wrapper(self):
+        sha = hashlib.sha256(NEW_BIN).hexdigest()
+        box = update_box(update_job(sha))
+        box.sig_rc = 1
+        rc, rep = run(box)
+        self.assertTrue(rep.get("refused"), f"a job whose signature does not verify was NOT refused: {rep}")
+        self.assertEqual((rc, rep["refused"]["code"]), (2, "R3"), rep)
+        self.assertEqual(wrapper_calls(box), [])
+
+    def test_a_staged_binary_the_signature_does_not_pin_is_refused(self):
+        sha = hashlib.sha256(NEW_BIN).hexdigest()
+        box = update_box(update_job(sha), content=b"\x7fELF something the agent put there")
+        rc, rep = run(box)
+        self.assertEqual((rc, rep["refused"]["code"]), (2, "R19"), rep)
+        self.assertEqual(wrapper_calls(box), [])
+        self.assertNotIn("u1", box.nonces, "a refused job keeps its nonce (the operator fixes the file, not the key)")
+
+    def test_another_staging_path_is_refused(self):
+        sha = hashlib.sha256(NEW_BIN).hexdigest()
+        box = update_box(update_job(sha), staged="/tmp/felhom-agent-0.146.0")
+        rc, rep = run(box)
+        self.assertEqual((rc, rep["refused"]["code"]), (2, "R19"), rep)
+        self.assertEqual(wrapper_calls(box), [])
+
+    def test_a_bundle_job_is_not_an_agent_update(self):
+        sha = hashlib.sha256(NEW_BIN).hexdigest()
+        box = update_box(update_job(sha, op="agent_config_update"))
+        rc, rep = run(box)
+        self.assertEqual((rc, rep["refused"]["code"]), (2, "R3"), rep)
+
+    def test_another_hosts_job_and_a_replay_are_refused(self):
+        sha = hashlib.sha256(NEW_BIN).hexdigest()
+        box = update_box(update_job(sha, host="tester-1-d70be4"))
+        self.assertEqual(run(box)[1]["refused"]["code"], "R3")
+        box2 = update_box(update_job(sha))
+        box2.nonces["u1"] = 1999999999
+        self.assertEqual(run(box2)[1]["refused"]["code"], "R3")
+        self.assertEqual(wrapper_calls(box) + wrapper_calls(box2), [])
+
+    def test_a_failed_flip_keeps_the_nonce(self):
+        sha = hashlib.sha256(NEW_BIN).hexdigest()
+        box = update_box(update_job(sha))
+        orig = box.host
+        box.host = lambda argv, timeout=600, stdin=None: (1, "", "refusing apply: sha mismatch") if argv[:1] == [osapply.SELFUPDATE_WRAPPER] else orig(argv, timeout, stdin)
+        rc, rep = run(box)
+        self.assertEqual(rc, 3, rep)
+        self.assertNotIn("u1", box.nonces)
+
+
+
+class SelfupdateWrapperConfinement(unittest.TestCase):
+    """R-861 (v0.146.1): the A/B wrapper takes only felhom-os-apply's root-owned copy, never the agent's staging dir
+    (a file there can be swapped between the wrapper's sha check and its copy). The path check runs before anything
+    is touched, so the real script can be run here unprivileged.
+    RED-PROOF: point ROOT_STAGING back at /var/lib/felhom-agent/selfupdate → this fails (the path is accepted and the
+    script goes on to `staged file missing`)."""
+
+    def test_the_agents_staging_dir_is_refused(self):
+        import subprocess
+        sha = "0" * 64
+        p = subprocess.run(["sh", str(HERE / "felhom-selfupdate-guarded"), "apply",
+                            "/var/lib/felhom-agent/selfupdate/felhom-agent-0.146.1", sha], capture_output=True, text=True)
+        self.assertEqual(p.returncode, 1, p.stderr)
+        self.assertIn("outside /var/lib/felhom-os-apply/agent-update", p.stderr)
+
+
+_sb = importlib.machinery.SourceFileLoader("stepbuild", str(REPO / "scripts" / "build-step-bundle.py"))
+_ss = importlib.util.spec_from_loader("stepbuild", _sb)
+stepbuild = importlib.util.module_from_spec(_ss)
+_sb.exec_module(stepbuild)
+NEW_IN_0146 = {"/usr/local/sbin/felhom-priv-apply", "/var/lib/vz/snippets/felhom-guest-hook.sh",
+               "/usr/local/sbin/felhom-shared-parent.sh", "/etc/systemd/system/felhom-shared-parent.service"}
+
+
+class StepBundle(unittest.TestCase):
+    """R-880 (agent v0.146.1): an INSTALLED wrapper checks an incoming bundle's paths against its OWN table (R16), so a
+    release that adds paths needs a step bundle: the boxes' current bundle with only felhom-os-apply replaced.
+    RED-PROOF: deliver the full bundle to the old table → R16 (test_the_full_bundle_is_refused_by_an_old_table)."""
+
+    def old_world(self):
+        """The base bundle an older wrapper (no R-861 paths) installed, and that wrapper's table."""
+        full = json.loads(builder.build("0.145.0"))
+        full["files"] = [e for e in full["files"] if e["path"] not in NEW_IN_0146]
+        old_wrapper = b'# the v0.145.0 wrapper stands in here\nBUNDLE_OP = "agent_config_update"\n'
+        for e in full["files"]:
+            if e["path"] == "/usr/local/sbin/felhom-os-apply":
+                e["content_b64"], e["sha256"] = base64.b64encode(old_wrapper).decode(), hashlib.sha256(old_wrapper).hexdigest()
+        base = (json.dumps(full, indent=1, sort_keys=True) + "\n").encode()
+        old_dests = {k: v for k, v in osapply.BUNDLE_DESTS.items() if k not in NEW_IN_0146}
+        return base, old_dests
+
+    def parse_with_table(self, data, dests, version):
+        saved = osapply.BUNDLE_DESTS
+        osapply.BUNDLE_DESTS = dests
+        try:
+            return osapply.Bundle(osapply.Apply(Box(b"{}", None), "")).parse(data, hashlib.sha256(data).hexdigest(), version)
+        finally:
+            osapply.BUNDLE_DESTS = saved
+
+    def test_the_full_bundle_is_refused_by_an_old_table(self):
+        _, old_dests = self.old_world()
+        full = builder.build("0.146.1")
+        with self.assertRaises(osapply.Refused) as cm:
+            self.parse_with_table(full, old_dests, "0.146.1")
+        self.assertEqual(cm.exception.code, "R16")
+
+    def test_the_step_bundle_is_accepted_by_the_old_table_and_changes_only_the_wrapper(self):
+        base, old_dests = self.old_world()
+        new_wrapper = (HERE / "felhom-os-apply").read_bytes()
+        step = stepbuild.build_step(base, "0.146.1-step1", new_wrapper)
+        ver, files = self.parse_with_table(step, old_dests, "0.146.1-step1")
+        self.assertEqual(ver, "0.146.1-step1")
+        b, s_ = json.loads(base), json.loads(step)
+        self.assertEqual(sorted(e["path"] for e in b["files"]), sorted(e["path"] for e in s_["files"]), "the paths must not change")
+        changed = [e["path"] for e, f in zip(sorted(b["files"], key=lambda x: x["path"]), sorted(s_["files"], key=lambda x: x["path"]))
+                   if e != f]
+        self.assertEqual(changed, ["/usr/local/sbin/felhom-os-apply"], "exactly the wrapper changes")
+        installed = dict((d, c) for d, c, *_ in files)
+        self.assertEqual(installed["/usr/local/sbin/felhom-os-apply"], new_wrapper)
+        # and the NEW wrapper (now installed) knows every path the release's full bundle names
+        self.assertTrue({e["path"] for e in json.loads(builder.build("0.146.1"))["files"]} <= set(osapply.BUNDLE_DESTS))
+
+    def test_a_step_version_must_carry_a_suffix(self):
+        base, _ = self.old_world()
+        with self.assertRaises(SystemExit):
+            stepbuild.build_step(base, "0.146.1", b"x")
+
 if __name__ == "__main__":
     unittest.main()

configs/test_felhom_os_apply.py

@@ -83,7 +83,7 @@ class Fake:
         return self.clock
 
     def sleep(self, s):
-        pass
+        self.sleeps = getattr(self, "sleeps", []) + [(len(self.calls), s)]  # (calls made before it, seconds)
 
     def verify_sig(self, signers, key_id, ns, blob, sig):
         self.verified = (signers, key_id, ns, blob, sig)
@@ -154,6 +154,8 @@ class Fake:
         if cmd == "dpkg" and a[1] == "--audit":
             return 0, self.dpkg_audit, ""
         if cmd == "dpkg" and a[1] == "--configure":
+            self.configured_calls = getattr(self, "configured_calls", 0) + 1
+            self.dpkg_journal = []  # `dpkg --configure -a` replays and empties the update journal
             return 0, "", ""
         if cmd == "fuser":
             return (0, " 123", "") if self.lock_held else (1, "", "")
@@ -176,6 +178,9 @@ class Fake:
             if "--print-uris" in a or "-s" in a:
                 return self.sim(a)  # --print-uris prints and installs nothing, with or without -s (9202, 2026-10-05)
             if "install" in a:
+                if getattr(self, "dpkg_journal", []):
+                    # real apt (demo-hp 2026-10-05): a non-empty update journal refuses every install
+                    return 100, "", "E: dpkg was interrupted, you must manually run 'sudo dpkg --configure -a' to correct the problem.\n"
                 if self.install_rc:
                     return self.install_rc, "", "E: boom"
                 for x in a:
@@ -191,6 +196,27 @@ class Fake:
             return 0, self.engine + "\n", ""
         if cmd == "docker" and a[1:3] == ["ps", "-q"]:
             return 0, "".join(i + "\n" for i in self.ids), ""
+        # R-528: the memory-kill check's engine (oom_image None = unreadable; oom_state / oom_event the engine's answer)
+        if cmd == "date" and a[1:] == ["+%s"]:
+            # each read is 3 s later than the last, so the order of the reads is visible in the values
+            self.oom_epochs = getattr(self, "oom_epochs", []) + [int(self.clock) + 3 * len(getattr(self, "oom_epochs", []))]
+            return 0, f"{self.oom_epochs[-1]}\n", ""
+        if cmd == "docker" and a[1:4] == ["inspect", "-f", "{{.Config.Image}}"]:
+            img = getattr(self, "oom_image", "gitea.dooplex.hu/admin/felhom-controller:0.300.0")
+            return (0, img + "\n", "") if img is not None else (1, "", "Error: No such object: felhom-controller")
+        if cmd == "docker" and a[1] == "run":
+            self.oom_runs = getattr(self, "oom_runs", []) + [a]
+            return 137, "", ""
+        if cmd == "docker" and a[1:4] == ["inspect", "-f", "{{.State.OOMKilled}} {{.State.ExitCode}}"]:
+            if getattr(self, "oom_inspect_raises", False):
+                raise subprocess.TimeoutExpired(a, 10)
+            return 0, getattr(self, "oom_state", "true 137") + "\n", ""
+        if cmd == "docker" and a[1] == "events":
+            self.oom_events_argv = a
+            return 0, ("oom\n" if getattr(self, "oom_event", True) else ""), ""
+        if cmd == "docker" and a[1:3] == ["rm", "-f"]:
+            self.oom_removed = getattr(self, "oom_removed", []) + a[3:]
+            return 0, a[3] + "\n", ""
         if cmd == "docker" and a[1] == "inspect":
             mounts = {"aaa111": "/felhom-controller|/var/run/docker.sock;/app/data;", "bbb222": "/app|/data;"}
             return 0, mounts.get(a[-1], "/other|;") + "\n", ""
@@ -220,6 +246,8 @@ class Fake:
             return 0, "", ""
         if cmd == "getent":
             return 0, "1.2.3.4 deb.debian.org\n", ""
+        if cmd == "sh" and a[2] == osapply.DPKG_STATE_SCRIPT:
+            return 0, self.dpkg_audit + osapply.JOURNAL_MARK + "\n" + "".join(j + "\n" for j in getattr(self, "dpkg_journal", [])), ""
         if cmd == "sh":
             if "vmlinuz" in a[2]:
                 return 0, "/boot/vmlinuz-7.0.2-6-pve\n/boot/vmlinuz-7.0.14-20-pve\n", ""
@@ -990,8 +1018,6 @@ class RealSignatureCheck(unittest.TestCase):
         self.assertNotEqual(r.verify_sig(self.signers, "someone-else", "felhom-op-v1", blob, sig), 0)
         self.assertNotEqual(r.verify_sig(self.signers, "felhom-op-1", "felhom-op-v1", blob, self.sign(blob, ns="other-ns")), 0)
 
-if __name__ == "__main__":
-    unittest.main()
 
 
 class UnsentReport(unittest.TestCase):
@@ -1077,3 +1103,226 @@ class SaveReportOnDisk(unittest.TestCase):
         with self.assertRaises(OSError):
             self.r.save_report(os.path.join(link, "plan-r3-guest-apply.json"), {"mode": "apply"})
         self.assertEqual(os.listdir(real), [])
+
+
+class AgentDiesMidPass(unittest.TestCase):
+    """R-868, MEASURED LIVE 2026-10-05 05:45 UTC on demo-hp (agent v0.144.0): the agent was kill -9-ed while apt-get
+    ran; apt finished all 13 packages, but the wrapper's next log line went to a stderr pipe nobody reads any more ->
+    BrokenPipeError -> the wrapper died before save_report: no DONE in the journal, no kept copy, no report.
+    COMPANION RED-PROOF: let Runner.log write to stderr unguarded -> this test fails (BrokenPipeError)."""
+
+    def test_a_dead_reader_does_not_stop_the_report_copy(self):
+        import io, sys, contextlib
+
+        class DeadPipe(io.TextIOBase):
+            dead = False
+            def write(self, s):
+                if DeadPipe.dead:
+                    raise BrokenPipeError(32, "Broken pipe")
+                return len(s)
+            def flush(self):
+                if DeadPipe.dead:
+                    raise BrokenPipeError(32, "Broken pipe")
+
+        class DyingFake(Fake):
+            def emulate(self, argv):
+                a = [x for x in argv if not re.match(r"^[A-Z_]+=", x) and x != "env"]
+                if a and a[0] == "apt-get" and "install" in a and "-s" not in a and "--print-uris" not in a:
+                    DeadPipe.dead = True  # the agent is killed while apt-get runs
+                return super().emulate(argv)
+
+        f = DyingFake()
+        journal = []
+        f.log = lambda line: osapply.Runner.log(f, line)  # the REAL log(): stderr, then the journal
+        prev_run = osapply.subprocess.run
+        osapply.subprocess.run = lambda argv, *a, **kw: (journal.append(argv[-1]) if argv[0] == "logger"
+                                                         else prev_run(argv, *a, **kw))  # never the real `logger`
+        prev_err, prev_out = sys.stderr, sys.stdout
+        sys.stderr, sys.stdout = DeadPipe(), DeadPipe()
+        try:
+            rc = osapply.main(["felhom-os-apply", "--plan", PLAN], runner=f)
+        finally:
+            sys.stderr, sys.stdout = prev_err, prev_out
+            osapply.subprocess.run = prev_run
+            DeadPipe.dead = False
+        self.assertEqual(rc, 0)
+        self.assertEqual(len(f.saved_reports), 1, "the kept copy is the only way this report reaches the hub")
+        self.assertEqual(len(f.saved_reports[0][1]["upgraded"]), 2)
+        self.assertTrue(any(l.startswith("os-apply: DONE") for l in journal), "the journal must still get the DONE line")
+
+
+class CrashLeftTheJournal(unittest.TestCase):
+    """R-876 — THE MEASURED SHAPE (demo-hp 2026-10-05, a crash while dpkg unpacked): `dpkg --audit` CLEAN, but
+    /var/lib/dpkg/updates holds 3 files; apt refuses every install ("dpkg was interrupted"). v0.144.1 logged
+    `REPAIR configured=0 fixed=0`, then `FAILED rc=100`, every pass, until a person ran `dpkg --configure -a`.
+    COMPANION RED-PROOFS: drop `or journal` from repair()'s condition -> test 1 fails; drop the interrupted-retry
+    -> test 3 fails; make repair() always run -> test 2 fails (R-845's speed)."""
+
+    def test_the_next_pass_repairs_by_itself_and_finishes(self):
+        f = Fake()
+        f.dpkg_audit = ""
+        f.dpkg_journal = ["0000", "0001", "0002"]
+        rc, rep = run(f)
+        self.assertEqual(rc, 0, rep)
+        self.assertEqual(rep["repair"]["journal_before"], 3)
+        self.assertTrue(rep["repair"]["clean_after"])
+        self.assertEqual(len(rep["upgraded"]), 2)
+        cfg = next(i for i, c in enumerate(f.calls) if c[0] == "guest" and "--configure" in c[2])
+        inst = next(i for i, c in enumerate(f.calls) if c[0] == "guest" and "install" in c[2] and "-s" not in c[2]
+                    and "-f" not in c[2] and "--print-uris" not in c[2])
+        self.assertLess(cfg, inst, "the repair must run before the install")
+        self.assertFalse(any("INTERRUPTED" in l for l in f.logs), "the journal check must catch it BEFORE apt refuses")
+
+    def test_a_clean_pass_still_costs_one_state_call(self):
+        f = Fake()
+        rc, rep = run(f)
+        self.assertEqual(rc, 0, rep)
+        state = [c for c in f.calls if c[0] == "guest" and c[2][:2] == ["sh", "-c"] and c[2][2] == osapply.DPKG_STATE_SCRIPT]
+        self.assertEqual(len(state), 1, "R-845: a clean pass reads dpkg's state ONCE and runs no repair")
+        self.assertEqual(getattr(f, "configured_calls", 0), 0)
+
+    def test_apt_interrupted_is_repaired_and_retried_once(self):
+        """The belt: the journal probe saw nothing (a race, an odd layout), apt still says interrupted."""
+        f = Fake()
+        orig = f.emulate
+        state = {"first": True}
+
+        def emulate(argv):
+            a = [x for x in argv if not re.match(r"^[A-Z_]+=", x) and x != "env"]
+            if a[0] == "apt-get" and "install" in a and "-s" not in a and "-f" not in a and "--print-uris" not in a and state["first"]:
+                state["first"] = False
+                return 100, "", "E: dpkg was interrupted, you must manually run 'sudo dpkg --configure -a' to correct the problem.\n"
+            return orig(argv)
+        f.emulate = emulate
+        rc, rep = run(f)
+        self.assertEqual(rc, 0, rep)
+        self.assertTrue(any("INTERRUPTED" in l for l in f.logs), f.logs)
+        self.assertTrue(any(l.startswith("os-apply: REPAIR ") and l.endswith("forced") for l in f.logs), f.logs)
+
+
+class OOMCheck(unittest.TestCase):
+    """R-528 (`09` decision 157): after a Docker engine step the wrapper proves the engine reports a memory kill
+    (OOMKilled=true AND the `oom` event). Reported only; the hub decides. Red-proofs: audits/readback-2026-10-07/F/."""
+
+    def apply(self, **kw):
+        f = docker_fake(signed=signed_job())
+        for k, v in kw.items():
+            setattr(f, k, v)
+        rc, rep = run(f)
+        self.assertEqual(rc, 0, rep)
+        return f, rep
+
+    def test_pass_when_oomkilled_and_the_event(self):
+        f, rep = self.apply()
+        oc = rep["oom_check"]
+        self.assertEqual(oc["result"], "pass", oc)
+        self.assertEqual((oc["oom_killed"], oc["oom_event"], oc["exit_code"]), (True, True, 137))
+        self.assertEqual(oc["image"], "gitea.dooplex.hu/admin/felhom-controller:0.300.0")
+        self.assertEqual(sorted(oc), ["detail", "exit_code", "image", "oom_event", "oom_killed", "result"])
+        run_argv = f.oom_runs[0]
+        name = run_argv[run_argv.index("--name") + 1]
+        self.assertTrue(re.match(r"^felhom-oomcheck-[0-9]+-[0-9a-f]{8}$", name), name)
+        for flag, val in (("--pull", "never"), ("--network", "none"), ("--memory", "64m"), ("--memory-swap", "64m"),
+                          ("--label", "felhom.oomcheck=1"), ("--entrypoint", "sh")):
+            self.assertEqual(run_argv[run_argv.index(flag) + 1], val, flag)
+        self.assertNotIn("-v", run_argv)
+        self.assertNotIn("-p", run_argv)
+        self.assertEqual(run_argv[-3:], ["gitea.dooplex.hu/admin/felhom-controller:0.300.0", "-c", osapply.OOMCHECK_SCRIPT])
+        self.assertIn(f"container={name}", f.oom_events_argv)
+        self.assertIn("event=oom", f.oom_events_argv)
+        self.assertEqual(f.oom_removed, [name], "the check container must be removed")
+        self.assertTrue(rep["health_after"], "health is read before the check")
+        # bounded: every call has a timeout and the clock is read twice — the worst case stays within 90 s
+        self.assertLessEqual(sum(osapply.OOMCHECK_TIMEOUTS.values()) + osapply.OOMCHECK_TIMEOUTS["clock"]
+                             + osapply.OOMCHECK_SETTLE, 90)
+
+    def test_events_window_ends_after_the_settle_wait(self):
+        # Measured (F1/F2): an --until taken right after the run missed the oom event. The window must end after a
+        # wait of >= 2 s that comes AFTER the run, at the guest epoch read after that wait, + 1.
+        f, rep = self.apply()
+        run_i = next(i for i, c in enumerate(f.calls) if c[-1][:2] == ["docker", "run"])
+        date_i = [i for i, c in enumerate(f.calls) if c[-1] == ["date", "+%s"]]
+        waits = [(i, s) for i, s in getattr(f, "sleeps", []) if i > run_i]
+        self.assertTrue(waits and waits[0][1] >= 2, f"no settle wait after the run: {getattr(f, 'sleeps', None)}")
+        self.assertTrue(date_i[-1] >= waits[0][0], "the end epoch must be read after the wait")
+        ev = f.oom_events_argv
+        since, until = int(ev[ev.index("--since") + 1]), int(ev[ev.index("--until") + 1])
+        self.assertEqual(until, f.oom_epochs[-1] + 1, "until = the guest epoch read after the wait, + 1")
+        self.assertGreater(until, f.oom_epochs[0] + 1)
+        self.assertLess(since, f.oom_epochs[0] + 1)
+
+    def test_oomkilled_false_is_fail(self):
+        f, rep = self.apply(oom_state="false 0")
+        oc = rep["oom_check"]
+        self.assertEqual(oc["result"], "fail", oc)
+        self.assertIn("OOMKilled=true", oc["detail"])
+        self.assertTrue(oc["oom_event"])
+        self.assertEqual(len(f.oom_removed), 1)
+
+    def test_no_event_is_fail(self):
+        f, rep = self.apply(oom_event=False)
+        oc = rep["oom_check"]
+        self.assertEqual(oc["result"], "fail", oc)
+        self.assertIn("the oom event", oc["detail"])
+        self.assertTrue(oc["oom_killed"])
+
+    def test_image_unreadable_is_error(self):
+        f, rep = self.apply(oom_image=None)
+        oc = rep["oom_check"]
+        self.assertEqual(oc["result"], "error", oc)
+        self.assertIsNone(oc["image"])
+        self.assertIn("image could not be read", oc["detail"])
+        self.assertFalse(hasattr(f, "oom_runs"), "no container is started without an image")
+
+    def test_container_removed_even_when_inspect_raises(self):
+        f, rep = self.apply(oom_inspect_raises=True)
+        oc = rep["oom_check"]
+        self.assertEqual(oc["result"], "error", oc)
+        self.assertEqual(len(f.oom_removed), 1, "the container must be removed even when inspect raised")
+        self.assertTrue(f.oom_removed[0].startswith(osapply.OOMCHECK_PREFIX))
+        self.assertNotIn("failed", rep, "the check never turns the step into a failure")
+
+    def test_never_on_guest_or_host_or_in_health_mode(self):
+        g = Fake()
+        rc, rep = run(g)
+        self.assertEqual(rc, 0, rep)
+        h = Fake()
+        h.plan["layer"] = "host"
+        h.plan["packages"] = [{"name": "bash", "version": "5.2.37-2+b10", "origin": "Debian"}]
+        h.installed["bash"] = "5.2.37-2+b9"
+        h.live["bash"] = {"5.2.37-2+b10"}
+        rc_h, rep_h = run(h)
+        self.assertEqual(rc_h, 0, rep_h)
+        d = docker_fake(signed=signed_job())
+        d.plan["mode"] = "health"
+        rc_d, rep_d = run(d)
+        self.assertEqual(rc_d, 0, rep_d)
+        for f, r in ((g, rep), (h, rep_h), (d, rep_d)):
+            self.assertNotIn("oom_check", r)
+            self.assertFalse(hasattr(f, "oom_runs"), r.get("layer"))
+            self.assertFalse(any(c[-1][:2] == ["docker", "run"] for c in f.calls))
+
+    def test_mode_oom_check_runs_only_the_check(self):
+        f = docker_fake()  # no authority: the check changes nothing, so it needs none
+        f.plan["mode"], f.plan["packages"] = "oom-check", []
+        rc, rep = run(f)
+        self.assertEqual(rc, 0, rep)
+        self.assertEqual(rep["oom_check"]["result"], "pass", rep)
+        self.assertFalse(any("apt-get" in c[-1] or "dpkg-query" in c[-1] for c in f.calls), f.calls)
+        self.assertEqual(len(f.oom_removed), 1)
+
+    def test_mode_oom_check_keeps_the_refusals(self):
+        f = docker_fake()
+        f.plan["mode"], f.plan["packages"] = "oom-check", []
+        f.files["/etc/pve/lxc/9201.conf"] = "arch: amd64\n"
+        rc, rep = run(f)
+        self.assertEqual((rc, rep["refused"]["code"]), (2, "R10"), rep)
+        self.assertFalse(hasattr(f, "oom_runs"))
+        g = Fake()
+        g.plan["mode"] = "oom-check"
+        rc, rep = run(g)
+        self.assertEqual((rc, rep["refused"]["code"]), (2, "R11"), rep)
+
+
+if __name__ == "__main__":
+    unittest.main()

configs/test_felhom_priv_apply.py

@@ -0,0 +1,320 @@
+#!/usr/bin/env python3
+"""Tests for felhom-priv-apply (R-861, `03` §3.1). An in-memory host plays the files; nothing real is written or run.
+Each refusal rule has a test that feeds it the ATTACK it exists for; each accepted shape is a file the agent really
+renders (the Go contract tests feed the live renderers too). Red-proof: audits/hub-safety-2026-10-05/partF/.
+
+Run: python3 configs/test_felhom_priv_apply.py   (also run by internal/privapply's Go test)
+"""
+import sys
+sys.dont_write_bytecode = True
+import importlib.machinery
+import importlib.util
+import os
+import pathlib
+import unittest
+
+HERE = pathlib.Path(__file__).resolve().parent
+_loader = importlib.machinery.SourceFileLoader("privapply", os.environ.get("PRIVAPPLY_UNDER_TEST", str(HERE / "felhom-priv-apply")))
+_spec = importlib.util.spec_from_loader("privapply", _loader)
+pa = importlib.util.module_from_spec(_spec)
+_loader.exec_module(pa)
+
+AGENT_UID = 999
+
+
+class FakeHost:
+    def __init__(self):
+        self.src, self.dest, self.logs, self.writes = {}, {}, [], []
+        self.owner, self.kind = {}, {}
+
+    def stage(self, path, text, uid=AGENT_UID, kind="file"):
+        self.src[path] = text.encode() if isinstance(text, str) else text
+        self.owner[path], self.kind[path] = uid, kind
+
+    def agent_uid(self):
+        return AGENT_UID
+
+    def read_source(self, path):
+        # mirrors Host.read_source's refusals: absent, not a regular file (a symlink is refused by O_NOFOLLOW), owner, size
+        if path not in self.src:
+            raise pa.Refused("P1", f"cannot open the staged file {path}")
+        if self.kind[path] != "file":
+            raise pa.Refused("P1", f"{path} is not a regular file")
+        if self.owner[path] != AGENT_UID:
+            raise pa.Refused("P1", f"{path} is not owned by felhom-agent")
+        if len(self.src[path]) > pa.MAX_BYTES:
+            raise pa.Refused("P1", f"{path} is larger than {pa.MAX_BYTES} bytes")
+        return self.src[path]
+
+    def read_dest(self, path):
+        return self.dest.get(path)
+
+    def install(self, dest, data, mode):
+        self.writes.append((dest, mode))
+        self.dest[dest] = data
+
+    def log(self, line):
+        self.logs.append(line)
+
+
+LOCAL_UNIT = """# Managed by felhom-agent — do not edit by hand.
+[Unit]
+Description=Felhom storage mount 91d2dc2d-2d28-4929-9bdd-3e11fa2f41ae
+After=local-fs-pre.target
+
+[Mount]
+What=/dev/disk/by-uuid/91d2dc2d-2d28-4929-9bdd-3e11fa2f41ae
+Where=/mnt/hdd_1
+Type=ext4
+
+[Install]
+WantedBy=multi-user.target
+"""
+NET_MOUNT = """# felhom network storage — do not edit by hand.
+[Unit]
+Description=Felhom network storage media (nfs)
+
+[Mount]
+What=nas.lan:/volume1/media
+Where=/mnt/felhom-drives/media
+Type=nfs4
+Options=vers=4.1,soft,timeo=50,retrans=2,noatime,_netdev,retry=0,nosuid,nodev
+"""
+NET_AUTOMOUNT = """# felhom network storage — do not edit by hand.
+[Unit]
+Description=Felhom network storage automount media (nfs)
+
+[Automount]
+Where=/mnt/felhom-drives/media
+TimeoutIdleSec=600
+
+[Install]
+WantedBy=multi-user.target
+"""
+NET_NAME = "mnt-felhom\\x2ddrives-media.mount"
+DNS_BASE = """# felhom split-horizon resolver — host base config (agent-managed; DO NOT EDIT)
+bind-interfaces
+listen-address=192.168.0.104
+listen-address=127.0.0.1
+no-resolv
+server=1.1.1.1
+server=9.9.9.9
+"""
+DNS_GUEST = """# felhom split-horizon DNS — customer demo-hp (agent-managed; DO NOT EDIT)
+local=/enkisfelhom.hu/
+address=/enkisfelhom.hu/192.168.0.138
+"""
+K = "AAECAwQFBgcICQoLDA0ODxAREhMUFRYXGBkaGxwdHh8="  # base64 of 32 bytes
+WG = f"""# felhom offsite tunnel — agent-managed (S3); DO NOT EDIT
+[Interface]
+PrivateKey = {K}
+Address = 10.77.0.3/32
+MTU = 1280
+
+[Peer]
+PublicKey = {K}
+Endpoint = 49.12.1.2:51820
+AllowedIPs = 10.77.0.1/32, 10.77.0.250/32
+PersistentKeepalive = 25
+"""
+KEYLINE = "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIL8z0qCNgA3x2xxAB0Qj5ro8waFjGZ8Ta/sWB63tlLw+ felhom-op-1\n"
+
+
+def run(host, *argv):
+    return pa.main(list(argv), host=host)
+
+
+class Accepts(unittest.TestCase):
+    """What the agent really writes is accepted and installed, root-owned, at the fixed destination."""
+
+    def test_local_unit_installed(self):
+        h = FakeHost()
+        h.stage("/var/lib/felhom-agent/units/mnt-hdd_1.mount", LOCAL_UNIT)
+        self.assertEqual(run(h, "unit", "mnt-hdd_1.mount"), 0)
+        self.assertEqual(h.writes, [("/etc/systemd/system/mnt-hdd_1.mount", 0o644)])
+
+    def test_local_unit_without_type(self):  # the N100's unit has no Type= (autodetect)
+        h = FakeHost()
+        h.stage("/var/lib/felhom-agent/units/mnt-hdd_1.mount", LOCAL_UNIT.replace("Type=ext4\n", ""))
+        self.assertEqual(run(h, "unit", "mnt-hdd_1.mount"), 0)
+
+    def test_network_pair(self):
+        h = FakeHost()
+        h.stage("/var/lib/felhom-agent/units/" + NET_NAME, NET_MOUNT)
+        h.stage("/var/lib/felhom-agent/units/mnt-felhom\\x2ddrives-media.automount", NET_AUTOMOUNT)
+        self.assertEqual(run(h, "unit", NET_NAME), 0)
+        self.assertEqual(run(h, "unit", "mnt-felhom\\x2ddrives-media.automount"), 0)
+
+    def test_identical_is_not_rewritten(self):
+        h = FakeHost()
+        h.stage("/var/lib/felhom-agent/units/mnt-hdd_1.mount", LOCAL_UNIT)
+        h.dest["/etc/systemd/system/mnt-hdd_1.mount"] = LOCAL_UNIT.encode()
+        self.assertEqual(run(h, "unit", "mnt-hdd_1.mount"), 0)
+        self.assertEqual(h.writes, [])
+
+    def test_dnsmasq_both_dropins(self):
+        h = FakeHost()
+        h.stage("/tmp/felhom-resolver-123.conf", DNS_BASE)
+        self.assertEqual(run(h, "dnsmasq", "/tmp/felhom-resolver-123.conf", "felhom-resolver-base.conf"), 0)
+        h.stage("/tmp/felhom-resolver-124.conf", DNS_GUEST)
+        self.assertEqual(run(h, "dnsmasq", "/tmp/felhom-resolver-124.conf", "felhom-demo-hp.conf"), 0)
+        self.assertIn(("/etc/dnsmasq.d/felhom-demo-hp.conf", 0o644), h.writes)
+
+    def test_wg_installed_0600(self):
+        h = FakeHost()
+        h.stage(pa.WG_SRC, WG)
+        self.assertEqual(run(h, "wg"), 0)
+        self.assertEqual(h.writes, [(pa.WG_DEST, 0o600)])
+
+    def test_sshd_template_and_key(self):
+        h = FakeHost()
+        h.stage(pa.SSHD_SRC, pa.render_sshd(8822))
+        h.stage(pa.KEY_SRC, KEYLINE)
+        self.assertEqual(run(h, "sshd-config"), 0)
+        self.assertEqual(run(h, "sshd-key"), 0)
+        h.stage(pa.KEY_SRC, "")  # clearing the operator login is allowed
+        self.assertEqual(run(h, "sshd-key"), 0)
+
+    def test_escape_matches_systemd(self):
+        self.assertEqual(pa.systemd_escape_path("/mnt/felhom-drives/media"), "mnt-felhom\\x2ddrives-media")
+        self.assertEqual(pa.systemd_escape_path("/mnt/hdd_1"), "mnt-hdd_1")
+        self.assertEqual(pa.systemd_escape_path("/mnt/.x"), "mnt-.x")  # a dot is escaped only at the very start
+
+
+class Refuses(unittest.TestCase):
+    """Each rule, with the attack it exists for. Nothing is written on a refusal."""
+
+    def refused(self, h, argv, rule):
+        rc = run(h, *argv)
+        self.assertIn(rc, (2, 3), f"{argv} was accepted")
+        self.assertEqual(h.writes, [], f"{argv} wrote something")
+        self.assertTrue(any(f"[{rule}]" in l for l in h.logs), f"{argv}: rule {rule} not logged: {h.logs}")
+
+    def unit(self, text, name="mnt-hdd_1.mount"):
+        h = FakeHost()
+        h.stage("/var/lib/felhom-agent/units/" + name, text)
+        return h, ["unit", name]
+
+    def test_U1_name_outside_mnt(self):
+        h = FakeHost()
+        self.refused(h, ["unit", "etc-sudoers.d.mount"], "U1")
+        self.refused(h, ["unit", "../../etc/x.mount"], "U1")
+        self.refused(h, ["unit", "mnt-x.service"], "U1")
+
+    def test_U2_service_section(self):
+        self.refused(*self.unit(LOCAL_UNIT + "\n[Service]\nExecStart=/bin/sh -c id\n"), "U2")
+
+    def test_U2_wants_starts_another_unit(self):  # review 2026-10-05: Wants=reboot.target would reboot the host
+        for extra in ("Wants=reboot.target", "Requires=felhom-agent-rollback.service", "Before=pve-guests.service"):
+            self.refused(*self.unit(LOCAL_UNIT.replace("After=local-fs-pre.target", "After=local-fs-pre.target\n" + extra)), "U2")
+        self.refused(*self.unit(LOCAL_UNIT.replace("After=local-fs-pre.target", "After=poweroff.target")), "U2")
+
+    def test_U2_continuation_line(self):
+        t = LOCAL_UNIT.replace("Description=Felhom storage mount 91d2dc2d-2d28-4929-9bdd-3e11fa2f41ae",
+                               "Description=Felhom storage mount \\")
+        self.refused(*self.unit(t), "U2")
+        self.refused(*self.unit(LOCAL_UNIT.replace("# Managed by felhom-agent", "# comment \\\n# Managed by felhom-agent")), "U2")
+
+    def test_U2_unknown_key(self):
+        self.refused(*self.unit(LOCAL_UNIT.replace("Type=ext4", "Type=ext4\nDirectoryMode=0777")), "U2")
+
+    def test_U3_bind_over_sudoers_dir(self):
+        # the R-861 attack: mount an agent-owned directory over /etc/sudoers.d
+        t = LOCAL_UNIT.replace("Where=/mnt/hdd_1", "Where=/etc/sudoers.d")
+        self.refused(*self.unit(t, "mnt-hdd_1.mount"), "U3")
+
+    def test_U3_name_must_match_where(self):
+        self.refused(*self.unit(LOCAL_UNIT.replace("Where=/mnt/hdd_1", "Where=/mnt/other")), "U3")
+
+    def test_U3_traversal_in_where(self):
+        # the name passes U1 and equals the escaped Where — ONLY the Where rule stops a mount at /mnt/../etc = /etc
+        self.assertEqual(pa.systemd_escape_path("/mnt/../etc") + ".mount", "mnt-..-etc.mount")
+        self.refused(*self.unit(LOCAL_UNIT.replace("Where=/mnt/hdd_1", "Where=/mnt/../etc"), "mnt-..-etc.mount"), "U3")
+
+    def test_U4_what_is_an_agent_directory(self):
+        t = LOCAL_UNIT.replace("What=/dev/disk/by-uuid/91d2dc2d-2d28-4929-9bdd-3e11fa2f41ae", "What=/var/lib/felhom-agent/evil")
+        self.refused(*self.unit(t), "U4")
+
+    def test_U4_tmpfs(self):
+        self.refused(*self.unit(LOCAL_UNIT.replace("Type=ext4", "Type=tmpfs")), "U4")
+
+    def test_U5_bind_option(self):
+        self.refused(*self.unit(LOCAL_UNIT.replace("Type=ext4", "Type=ext4\nOptions=bind")), "U5")
+
+    def test_U5_network_without_nosuid(self):
+        t = NET_MOUNT.replace(",nosuid,nodev", "")
+        self.refused(*self.unit(t, NET_NAME), "U5")
+
+    def test_U5_suid_option(self):
+        self.refused(*self.unit(LOCAL_UNIT.replace("Type=ext4", "Type=ext4\nOptions=suid,dev")), "U5")
+
+    def test_U3_network_outside_drives(self):
+        t = NET_MOUNT.replace("/mnt/felhom-drives/media", "/mnt/media")
+        self.refused(*self.unit(t, "mnt-media.mount"), "U3")
+
+    def test_D1_dhcp_script(self):  # runs as root
+        h = FakeHost()
+        h.stage("/tmp/felhom-resolver-1.conf", DNS_BASE + "dhcp-script=/var/lib/felhom-agent/x.sh\n")
+        self.refused(h, ["dnsmasq", "/tmp/felhom-resolver-1.conf", "felhom-x.conf"], "D1")
+
+    def test_D1_conf_dir_and_log_file(self):
+        for extra in ("conf-dir=/var/lib/felhom-agent\n", "log-facility=/etc/sudoers.d/x\n", "user=root\n"):
+            h = FakeHost()
+            h.stage("/tmp/felhom-resolver-1.conf", DNS_GUEST + extra)
+            self.refused(h, ["dnsmasq", "/tmp/felhom-resolver-1.conf", "felhom-x.conf"], "D1")
+
+    def test_D2_paths(self):
+        h = FakeHost()
+        self.refused(h, ["dnsmasq", "/etc/shadow", "felhom-x.conf"], "D2")
+        self.refused(h, ["dnsmasq", "/tmp/felhom-resolver-1.conf", "../sudoers.d/x.conf"], "D2")
+
+    def test_W1_postup(self):  # wg-quick runs PostUp as root
+        h = FakeHost()
+        h.stage(pa.WG_SRC, WG.replace("MTU = 1280", "MTU = 1280\nPostUp = /bin/sh -c id"))
+        self.refused(h, ["wg"], "W1")
+
+    def test_W2_values(self):
+        h = FakeHost()
+        h.stage(pa.WG_SRC, WG.replace("AllowedIPs = 10.77.0.1/32, 10.77.0.250/32", "AllowedIPs = 0.0.0.0/0"))
+        self.refused(h, ["wg"], "W2")
+
+    def test_S1_sshd_strictmodes(self):  # an AuthorizedKeysFile the agent owns + StrictModes no = root login
+        h = FakeHost()
+        h.stage(pa.SSHD_SRC, pa.render_sshd(8822) + "StrictModes no\n")
+        self.refused(h, ["sshd-config"], "S1")
+        h2 = FakeHost()
+        h2.stage(pa.SSHD_SRC, pa.render_sshd(8822).replace("/etc/felhom-sshd/authorized_keys/%u", "/var/lib/felhom-agent/k"))
+        self.refused(h2, ["sshd-config"], "S1")
+
+    def test_S1_port_22(self):
+        h = FakeHost()
+        h.stage(pa.SSHD_SRC, pa.render_sshd(22))
+        self.refused(h, ["sshd-config"], "S1")
+
+    def test_S2_key_options_and_two_keys(self):
+        h = FakeHost()
+        h.stage(pa.KEY_SRC, 'command="/bin/sh" ' + KEYLINE)
+        self.refused(h, ["sshd-key"], "S2")
+        h2 = FakeHost()
+        h2.stage(pa.KEY_SRC, KEYLINE + KEYLINE)
+        self.refused(h2, ["sshd-key"], "S2")
+
+    def test_P1_symlink_owner_size(self):
+        h = FakeHost()
+        h.stage(pa.WG_SRC, WG, kind="symlink")
+        self.refused(h, ["wg"], "P1")
+        h2 = FakeHost()
+        h2.stage(pa.WG_SRC, WG, uid=0)
+        self.refused(h2, ["wg"], "P1")
+        h3 = FakeHost()
+        h3.stage(pa.WG_SRC, "#" * (pa.MAX_BYTES + 1))
+        self.refused(h3, ["wg"], "P1")
+
+    def test_A1_usage(self):
+        h = FakeHost()
+        self.refused(h, ["install", "/etc/shadow"], "A1")
+        self.refused(h, ["wg", "/etc/shadow"], "A1")
+
+
+if __name__ == "__main__":
+    unittest.main(verbosity=2)

internal/backup/archive_completeness_test.go

@@ -241,3 +241,23 @@ func TestNewestArchiveTime_DistinctPhantomsEachAnnounced(t *testing.T) {
 		t.Errorf("got %d rejection lines for 2 distinct phantoms across 3 polls, want 2:\n%s", n, buf.String())
 	}
 }
+
+// R-99 (`09` §3 decision 140): the WARN for a PBS phantom ends with the cleanup runbook, so whoever sees it knows the
+// one sanctioned way to remove it; a tiny archive on a dir storage is not a PBS phantom and gets no pointer.
+// RED-PROOF: drop the `msg += phantomCleanupPointer` line → "the PBS phantom WARN does not end with the runbook pointer".
+func TestRejectedArchiveWarnNamesTheCleanupRunbook(t *testing.T) {
+	var buf bytes.Buffer
+	r := runnerWithContent(t, &buf, []proxmox.StorageContent{phantomEntry(), goodPBSEntry()})
+	if _, _, err := r.NewestArchiveTime(context.Background(), 9201); err != nil {
+		t.Fatal(err)
+	}
+	const want = "INCOMPLETE archive when computing tier freshness — it is not a successful backup — a phantom leftover; delete it by felhom.eu documentation/runbooks/pbs-phantom-cleanup.md (09 §3 decision 140)"
+	if !strings.Contains(buf.String(), want) {
+		t.Errorf("the PBS phantom WARN does not end with the runbook pointer:\n%s", buf.String())
+	}
+	local := phantomEntry()
+	local.Format, local.VolID = "tar.zst", "local:backup/vzdump-lxc-9201-2026_07_28-05_31_14.tar.zst"
+	if got := rejectedArchiveMessage(local); strings.Contains(got, "pbs-phantom-cleanup") {
+		t.Errorf("a dir-storage archive got the PBS runbook pointer: %s", got)
+	}
+}

internal/backup/backup_state.go

@@ -0,0 +1,131 @@
+package backup
+
+import (
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"sort"
+	"strconv"
+	"sync"
+	"time"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/hub"
+)
+
+// BackupSuccessState persists the newest SUCCESSFUL whole-guest backup per tier and guest (R-894).
+//
+// Why it exists. The due-check (`localapi` handleBackupDue) asks the tier's storage when a backup last
+// landed (R-84) and falls back to the in-memory record when the storage cannot be read. The in-memory
+// record is empty after an agent restart (Store, R-348), so "storage unreadable" right after a restart
+// read as "no record — DUE". Measured 2026-10-05 on demo-hp: the agent restarted at 04:57, the off-site
+// storage answered "Can't connect" at 06:25, the 7-day tier — last copy 2026-10-01 — read DUE, the
+// controller asked, and vzdump failed. This file is the last known copy the fallback reads instead.
+//
+// It is read ONLY when the storage cannot be read. A storage that answers is the ground truth and wins,
+// in both directions: an archive found there counts, and an archive absent there is absent even when
+// this file remembers a success (a pruned or deleted archive must make the tier due — the same reason
+// R-84 chose the storage over a persisted record). Pinned by
+// TestBackupDue_R894_SavedCopyIgnoredWhenStorageAnswers.
+//
+// Only SUCCESSES are written (the RestoreTestState rule): a failure must stay due and be retried, so a
+// record of a failure has no reader.
+type BackupSuccessState struct {
+	path string
+	mu   sync.Mutex
+	last map[string]savedSuccess // key(target, vmid) → the newest success
+}
+
+type savedSuccess struct {
+	target string
+	vmid   int
+	at     time.Time
+}
+
+// backupSuccessJSON is one entry on disk.
+type backupSuccessJSON struct {
+	Target    string `json:"target"`
+	VMID      int    `json:"vmid"`
+	StartedAt string `json:"started_at"`
+}
+
+func backupStateKey(target string, vmid int) string { return target + "/" + strconv.Itoa(vmid) }
+
+// NewBackupSuccessState opens (or creates) the state at path. A missing or unreadable file degrades to
+// "nothing known" — the pre-R-894 behaviour, which is DUE — and never wedges the daemon.
+func NewBackupSuccessState(path string) *BackupSuccessState {
+	s := &BackupSuccessState{path: path, last: map[string]savedSuccess{}}
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return s
+	}
+	var entries []backupSuccessJSON
+	if json.Unmarshal(data, &entries) != nil {
+		return s
+	}
+	for _, e := range entries {
+		t, perr := time.Parse(time.RFC3339, e.StartedAt)
+		if perr != nil {
+			continue // one unreadable entry must not lose the others
+		}
+		s.last[backupStateKey(e.Target, e.VMID)] = savedSuccess{target: e.Target, vmid: e.VMID, at: t.UTC()}
+	}
+	return s
+}
+
+// RecordBackupSuccess saves b when it is a success newer than the one on file. target is the tier the
+// job ran on (the due-check's key); a failure or an unparseable time is ignored.
+func (s *BackupSuccessState) RecordBackupSuccess(target string, b hub.Backup) error {
+	if s == nil || !b.Success {
+		return nil
+	}
+	t, err := time.Parse(time.RFC3339, b.StartedAt)
+	if err != nil {
+		return nil
+	}
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	k := backupStateKey(target, b.VMID)
+	if old, ok := s.last[k]; ok && !t.After(old.at) {
+		return nil
+	}
+	s.last[k] = savedSuccess{target: target, vmid: b.VMID, at: t.UTC()}
+	return s.saveLocked()
+}
+
+// LastKnownSuccess returns the newest saved success for this tier and guest (ok=false = none on file).
+func (s *BackupSuccessState) LastKnownSuccess(target string, vmid int) (time.Time, bool) {
+	if s == nil {
+		return time.Time{}, false
+	}
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	e, ok := s.last[backupStateKey(target, vmid)]
+	return e.at, ok
+}
+
+func (s *BackupSuccessState) saveLocked() error {
+	entries := make([]backupSuccessJSON, 0, len(s.last))
+	for _, e := range s.last {
+		entries = append(entries, backupSuccessJSON{Target: e.target, VMID: e.vmid, StartedAt: e.at.Format(time.RFC3339)})
+	}
+	// Deterministic file content (Go's map order is random).
+	sort.Slice(entries, func(i, j int) bool {
+		if entries[i].Target != entries[j].Target {
+			return entries[i].Target < entries[j].Target
+		}
+		return entries[i].VMID < entries[j].VMID
+	})
+	data, err := json.MarshalIndent(entries, "", "  ")
+	if err != nil {
+		return err
+	}
+	if err := os.MkdirAll(filepath.Dir(s.path), 0o755); err != nil {
+		return err
+	}
+	tmp := s.path + ".tmp"
+	if err := os.WriteFile(tmp, data, 0o600); err != nil {
+		os.Remove(tmp)
+		return err
+	}
+	return os.Rename(tmp, s.path)
+}

internal/backup/backup_state_test.go

@@ -0,0 +1,59 @@
+package backup
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/hub"
+)
+
+// R-894: the on-disk newest success per tier survives a restart (a new state from the same file).
+func TestBackupSuccessState_SurvivesRestart(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "backup-success-state.json")
+	s := NewBackupSuccessState(path)
+	at := time.Date(2026, 10, 1, 20, 15, 0, 0, time.UTC)
+	if err := s.RecordBackupSuccess("felhom-pbs", hub.Backup{VMID: 9201, Success: true, StartedAt: at.Format(time.RFC3339)}); err != nil {
+		t.Fatal(err)
+	}
+	got, ok := NewBackupSuccessState(path).LastKnownSuccess("felhom-pbs", 9201)
+	if !ok || !got.Equal(at) {
+		t.Fatalf("after a restart the saved copy must read back; got %v ok=%v", got, ok)
+	}
+	if _, ok := NewBackupSuccessState(path).LastKnownSuccess("local", 9201); ok {
+		t.Fatal("another tier must not borrow this tier's copy")
+	}
+}
+
+// Only a NEWER success replaces the saved one; failures and unparseable times are ignored.
+func TestBackupSuccessState_KeepsNewestSuccessOnly(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "s.json")
+	s := NewBackupSuccessState(path)
+	newer := time.Date(2026, 10, 5, 0, 0, 0, 0, time.UTC)
+	older := newer.Add(-48 * time.Hour)
+	for _, b := range []hub.Backup{
+		{VMID: 1, Success: true, StartedAt: newer.Format(time.RFC3339)},
+		{VMID: 1, Success: true, StartedAt: older.Format(time.RFC3339)},                 // older: ignored
+		{VMID: 1, Success: false, StartedAt: newer.Add(time.Hour).Format(time.RFC3339)}, // failure: ignored
+		{VMID: 1, Success: true, StartedAt: "not-a-time"},                               // unparseable: ignored
+	} {
+		if err := s.RecordBackupSuccess("t", b); err != nil {
+			t.Fatal(err)
+		}
+	}
+	if got, _ := NewBackupSuccessState(path).LastKnownSuccess("t", 1); !got.Equal(newer) {
+		t.Fatalf("want the newest success %v, got %v", newer, got)
+	}
+}
+
+// A corrupt file degrades to "nothing known" (the pre-R-894 DUE answer), never a crash.
+func TestBackupSuccessState_CorruptFileIsEmpty(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "s.json")
+	if err := os.WriteFile(path, []byte("{not json"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	if _, ok := NewBackupSuccessState(path).LastKnownSuccess("t", 1); ok {
+		t.Fatal("a corrupt file must read as nothing known")
+	}
+}

internal/backup/r874_first_eval_test.go

@@ -0,0 +1,63 @@
+package backup
+
+import (
+	"context"
+	"fmt"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/reconcile"
+)
+
+// R-874 (v0.145.0). THE MEASURED SHAPE (Part F spike, Tester 2): power-on sessions of ~1.5 h and ~5 min against a
+// 6 h evaluation ticker that restarts at every start — no restore-test ever evaluated. Now the first evaluation runs
+// FirstEval after start.
+// COMPANION RED-PROOF: drop the first-evaluation timer in Run (back to the bare ticker) → "no evaluation within".
+func TestR874_FirstEvaluationAfterStart(t *testing.T) {
+	var picks int32
+	s := NewScheduler(SchedulerOptions{
+		Runner: &fakeRTRunner{res: reconcile.RestoreTestResult{Pass: true, Verified: "boot+running"}},
+		Pick: func(context.Context) (string, error) {
+			atomic.AddInt32(&picks, 1)
+			return fmt.Sprintf("local:backup/vzdump-lxc-9201-%d.tar.zst", atomic.LoadInt32(&picks)), nil
+		},
+		Store: NewStore(), Spec: (&specSpy{}).build,
+		Cadence: 6 * time.Hour, FirstEval: 30 * time.Millisecond, Logger: quiet(),
+	})
+	ctx, cancel := context.WithCancel(context.Background())
+	done := make(chan struct{})
+	go func() { _ = s.Run(ctx); close(done) }()
+	deadline := time.Now().Add(3 * time.Second)
+	for atomic.LoadInt32(&picks) == 0 && time.Now().Before(deadline) {
+		time.Sleep(10 * time.Millisecond)
+	}
+	cancel()
+	<-done
+	if atomic.LoadInt32(&picks) == 0 {
+		t.Fatal("no evaluation within 3 s of start (FirstEval 30 ms) — a box with short sessions never gets a restore-test")
+	}
+}
+
+// The earned restraint stays: an agent that restarts before FirstEval never evaluates (a crash loop does not
+// hammer a failing tier).
+func TestR874_CrashLoopNeverEvaluates(t *testing.T) {
+	var picks int32
+	for i := 0; i < 5; i++ { // five quick "restarts"
+		s := NewScheduler(SchedulerOptions{
+			Runner: &fakeRTRunner{res: reconcile.RestoreTestResult{Pass: true}},
+			Pick:   func(context.Context) (string, error) { atomic.AddInt32(&picks, 1); return "x", nil },
+			Store:  NewStore(), Spec: (&specSpy{}).build,
+			Cadence: 6 * time.Hour, FirstEval: 200 * time.Millisecond, Logger: quiet(),
+		})
+		ctx, cancel := context.WithTimeout(context.Background(), 20*time.Millisecond)
+		_ = s.Run(ctx)
+		cancel()
+	}
+	if n := atomic.LoadInt32(&picks); n != 0 {
+		t.Fatalf("a restart before FirstEval evaluated %d time(s)", n)
+	}
+	if DefaultFirstEval != 30*time.Minute {
+		t.Fatalf("DefaultFirstEval = %s, the documented 30 min", DefaultFirstEval)
+	}
+}

internal/backup/runner.go

@@ -518,10 +518,25 @@ func (r *BackupRunner) warnRejectedArchiveOnce(e proxmox.StorageContent, why str
 	if seen {
 		return
 	}
-	r.logger.Warn("backup: ignoring an INCOMPLETE archive when computing tier freshness — it is not a successful backup",
+	r.logger.Warn(rejectedArchiveMessage(e),
 		"target", r.target, "vmid", e.VMID, "volid", e.VolID, "size_bytes", e.Size, "reason", why)
 }
 
+// phantomCleanupPointer names the runbook that removes a PBS phantom (R-99, `09` §3 decision 140: a leftover of an
+// aborted upload is deleted on the backup server, by a runbook, when one is seen — never automatically).
+const phantomCleanupPointer = " — a phantom leftover; delete it by felhom.eu documentation/runbooks/pbs-phantom-cleanup.md (09 §3 decision 140)"
+
+// rejectedArchiveMessage is the WARN text for a rejected archive. Only a PBS entry (format pbs-ct / pbs-vm) gets the
+// cleanup pointer: the runbook deletes on a PBS datastore, and a tiny archive on a dir storage is not a PBS phantom.
+// Pinned by TestRejectedArchiveWarnNamesTheCleanupRunbook.
+func rejectedArchiveMessage(e proxmox.StorageContent) string {
+	msg := "backup: ignoring an INCOMPLETE archive when computing tier freshness — it is not a successful backup"
+	if strings.HasPrefix(e.Format, "pbs-") {
+		msg += phantomCleanupPointer
+	}
+	return msg
+}
+
 // demo-felhom in a single afternoon of deploys (2026-07-26).
 //
 // Asking the STORAGE rather than persisting the store is deliberate:

internal/backup/schedule.go

@@ -60,12 +60,21 @@ type Scheduler struct {
 
 	// R-85 tier rotation. All optional: without them the scheduler behaves exactly as before
 	// (single tier via `pick`), which keeps every existing caller and test working untouched.
-	tiers    []string          // configured tier target ids, primary first
-	tierPick TierPicker        // newest archive on a named tier
-	rtState  *RestoreTestState // persisted last-successful-per-tier (drives oldest-first)
-	inFlight *InFlight         // shared with the backup path — Scenario F
+	tiers     []string          // configured tier target ids, primary first
+	tierPick  TierPicker        // newest archive on a named tier
+	rtState   *RestoreTestState // persisted last-successful-per-tier (drives oldest-first)
+	inFlight  *InFlight         // shared with the backup path — Scenario F
+	firstEval time.Duration     // R-874: the first evaluation after start
 }
 
+// DefaultFirstEval (R-874): the first due-ness evaluation runs 30 minutes after the agent starts, then every
+// cadence. MEASURED need (2026-10-05 Part F spike): a box whose power-on sessions are all shorter than the 6 h
+// interval (Tester 2: ~1.5 h and ~5 min) NEVER evaluated, because the ticker restarts at each start. 30 minutes
+// keeps the earned restraint below — a crash-looping agent restarts far more often than that and still never
+// evaluates — while a box that stays on for half an hour gets its due test. Pinned by
+// TestR874_FirstEvaluationAfterStart and TestR874_CrashLoopNeverEvaluates.
+const DefaultFirstEval = 30 * time.Minute
+
 // SchedulerOptions configures a Scheduler.
 type SchedulerOptions struct {
 	Runner RestoreTestRunner
@@ -81,6 +90,8 @@ type SchedulerOptions struct {
 	// 0 → no settle requirement (any archive is a candidate).
 	Settle time.Duration
 	Logger *slog.Logger
+	// FirstEval (R-874, v0.145.0) is when the FIRST evaluation runs after start; 0 → DefaultFirstEval.
+	FirstEval time.Duration
 
 	// R-85 (all optional — omit for the pre-R-85 single-tier behaviour):
 	// Tiers are the configured tier target ids (primary first); TierPick resolves an archive on a
@@ -110,6 +121,12 @@ func NewScheduler(opts SchedulerOptions) *Scheduler {
 		tierPick: opts.TierPick,
 		rtState:  opts.State,
 		inFlight: opts.InFlight,
+		firstEval: func() time.Duration {
+			if opts.FirstEval > 0 {
+				return opts.FirstEval
+			}
+			return DefaultFirstEval
+		}(),
 	}
 }
 
@@ -120,8 +137,9 @@ func NewScheduler(opts SchedulerOptions) *Scheduler {
 // trigger any more: its phase is the process's uptime, and agent deploys reset it, which is exactly
 // the defect R-86 removes. What decides that a test happens is `EvaluateDue`.
 //
-// It still does NOT evaluate immediately on start — the first evaluation is one interval in. That
-// is an EARNED restraint, kept deliberately: a restore is heavy, agent restarts are routine, and a
+// It still does NOT evaluate immediately on start. v0.145.0 (R-874): the first evaluation is
+// firstEval (30 min) in, then every interval — it was one full interval in, which a box with short
+// power-on sessions never reached. The restraint itself is EARNED and kept: a restore is heavy, agent restarts are routine, and a
 // crash-loop that evaluated at start would hammer a permanently-failing tier as fast as it could
 // restart. Due-ness does not expire while we wait, so the only cost is up to one interval of
 // latency on a tier that just became due. On-demand runs use `--selftest=restore-test`.
@@ -135,6 +153,16 @@ func (s *Scheduler) Run(ctx context.Context) error {
 	}
 	s.logger.Info("backup: restore-test scheduler starting (per-archive due-check)",
 		"eval_interval", s.cadence, "settle", s.settle)
+	first := time.NewTimer(s.firstEval)
+	defer first.Stop()
+	select {
+	case <-ctx.Done():
+		s.logger.Info("backup: restore-test scheduler shutting down", "reason", ctx.Err())
+		return nil
+	case <-first.C:
+		s.logger.Info("backup: restore-test first evaluation after start (R-874)", "after", s.firstEval)
+		s.tick(ctx)
+	}
 	t := time.NewTicker(s.cadence)
 	defer t.Stop()
 	for {

internal/backup/store.go

@@ -25,7 +25,14 @@ import (
 // merges the two — see hub.ProvenRestoreTestReporter. This store remains the ONLY place a FAILURE is
 // recorded, and that asymmetry is deliberate: a failing tier stays due and is retried, so a lost
 // failure heals itself, while a lost success leaves the system quietly less tested than it believes.
-// Backups are unaffected — their freshness has a ground truth on the storage (R-84).
+// Backups are NOT unaffected (corrected 2026-10-05, R-348): byTarget is in memory too, so after a restart the
+// reported backup LIST reads 0 until the next backup of each tier runs (daily local, weekly offsite) — measured
+// 2026-08-20, two consecutive host-reports with `0 backups` while `pvesm list` showed archives on both tiers. What
+// is unaffected is the hub's VERDICT: it looks back 7 days over stored reports (felhom.eu hub/internal/monitor/
+// deadline.go backupEvidenceLookback) and the storage stays the ground truth (R-84).
+// The due-check's fallback for an UNREADABLE storage no longer reads this store alone (R-894): the newest
+// success per tier is also on disk (BackupSuccessState), so a restart followed by an unreachable storage
+// reads the last known copy, not "never".
 type Store struct {
 	mu       sync.Mutex
 	byTarget map[string]hub.Backup // latest backup per target id

internal/capability/manifest.go

@@ -81,10 +81,8 @@ var manifest = []Capability{
 	{"drives-mkdir-sub", "per-drive stable dir create", "/usr/bin/mkdir", []string{"-p", "/mnt/felhom-drives/felhom-usb"}, false, ""},
 	{"drives-mkdir-data", "felhom-data namespace create", "/usr/bin/mkdir", []string{"-p", "/mnt/felhom-usb/felhom-data"}, false, ""},
 	{"drives-chown-data", "felhom-data guest-root chown", "/usr/bin/chown", []string{"100000:100000", "/mnt/felhom-usb/felhom-data"}, false, ""},
-	{"parent-script-install", "shared-parent boot script install", "/usr/bin/install", []string{"-m", "0755", "--", "/tmp/felhom-shared-parent-123456789.sh", "/usr/local/sbin/felhom-shared-parent.sh"}, false, ""},
-	{"parent-unit-install", "shared-parent boot unit install", "/usr/bin/install", []string{"-m", "0644", "--", "/tmp/felhom-shared-parent-123456789.service", "/etc/systemd/system/felhom-shared-parent.service"}, false, ""},
 	{"parent-unit-enable", "shared-parent boot-persistence enable", "/usr/bin/systemctl", []string{"enable", "felhom-shared-parent.service"}, false, ""},
-	{"parent-bind-mp8", "parent bind into guest at provision", "/usr/sbin/pct", []string{"set", "9201", "-mp8", "/mnt/felhom-drives"}, false, ""},
+	{"parent-bind-mp8", "parent bind into guest at provision", "/usr/sbin/pct", []string{"set", "9201", "-mp8", "/mnt/felhom-drives,mp=/mnt/felhom-drives"}, false, ""},
 
 	// ---- Disk inspect / format gate (Critical: the data-bearing classifier + format) ----
 	{"disk-blkid", "disk data-bearing classify (format gate)", "/usr/sbin/blkid", []string{"-p", "-o", "export", "/dev/sda"}, true, ""},
@@ -95,11 +93,11 @@ var manifest = []Capability{
 	{"disk-lvs", "thin-pool usage read", "/usr/sbin/lvs", []string{"--reportformat", "json", "--units", "b", "-o", "lv_name,data_percent,metadata_percent", "--", "pve/data"}, false, ""},
 
 	// ---- Storage mount units (watchdog re-mount) ----
-	{"mount-unit-install", "fs-UUID mount unit install", "/usr/bin/install", []string{"-o", "root", "-g", "root", "-m", "0644", "--", "/var/lib/felhom-agent/units/felhom-x.mount", "/etc/systemd/system/felhom-x.mount"}, false, ""},
+	{"mount-unit-install", "fs-UUID mount unit install (root content check, R-861)", "/usr/local/sbin/felhom-priv-apply", []string{"unit", "mnt-felhom\\x2dx.mount"}, false, ""},
 	{"mount-daemon-reload", "systemd reload after unit write", "/usr/bin/systemctl", []string{"daemon-reload"}, false, ""},
-	{"mount-unit-enable", "mount unit enable", "/usr/bin/systemctl", []string{"enable", "--now", "--", "felhom-x.mount"}, false, ""},
-	{"mount-unit-disable", "mount unit disable", "/usr/bin/systemctl", []string{"disable", "--", "felhom-x.mount"}, false, ""},
-	{"mount-unit-stop", "mount unit stop", "/usr/bin/systemctl", []string{"stop", "--", "felhom-x.mount"}, false, ""},
+	{"mount-unit-enable", "mount unit enable", "/usr/bin/systemctl", []string{"enable", "--now", "--", "mnt-felhom\\x2dx.mount"}, false, ""},
+	{"mount-unit-disable", "mount unit disable", "/usr/bin/systemctl", []string{"disable", "--", "mnt-felhom\\x2dx.mount"}, false, ""},
+	{"mount-unit-stop", "mount unit stop", "/usr/bin/systemctl", []string{"stop", "--", "mnt-felhom\\x2dx.mount"}, false, ""},
 
 	// ---- Network storage re-arm + cleanup (CAMPAIGN-3 F10/F1) ----
 	{"netmount-reset-failed", "NAS automount re-arm after start-limit (F10)", "/usr/bin/systemctl", []string{"reset-failed", "--", "mnt-felhom\\x2ddrives-media.automount"}, false, ""},
@@ -110,18 +108,17 @@ var manifest = []Capability{
 
 	// ---- Provisioning back-half ----
 	{"provision-chown", "bootstrap mount guest-root chown", "/usr/bin/chown", []string{"-R", "100000:100000", "/var/lib/felhom-agent/guests/9201"}, false, ""},
-	{"provision-config-mount", "bootstrap config bind mount", "/usr/sbin/pct", []string{"set", "9201", "-mp0", "/var/lib/felhom-agent/guests/9201"}, false, ""},
+	{"provision-config-mount", "bootstrap config bind mount", "/usr/sbin/pct", []string{"set", "9201", "-mp0", "/var/lib/felhom-agent/guests/9201/bootstrap,mp=/etc/felhom-bootstrap,ro=1"}, false, ""},
 	{"provision-onboot", "customer guest autostart (onboot)", "/usr/sbin/pct", []string{"set", "9201", "-onboot", "1"}, false, ""},
 
 	// ---- Pre-start self-heal hook + guest lifecycle ----
-	{"guesthook-install", "pre-start hook snippet install", "/usr/bin/install", []string{"-m", "0755", "--", "/tmp/felhom-guest-hook-123456789.sh", "/var/lib/vz/snippets/felhom-guest-hook.sh"}, false, ""},
 	{"guesthook-register", "pre-start hook register", "/usr/sbin/pct", []string{"set", "9201", "--hookscript", "local:snippets/felhom-guest-hook.sh"}, false, ""},
 	{"guesthook-delete-mp", "dead mountpoint slot delete (C1 net)", "/usr/sbin/pct", []string{"set", "9201", "--delete", "mp0"}, false, ""},
 	{"guest-reboot", "enroll activate-binds reboot", "/usr/sbin/pct", []string{"reboot", "9201"}, false, ""},
 
 	// ---- LAN split-horizon resolver (dnsmasq) ----
 	{"dnsmasq-install", "dnsmasq package install", "/usr/bin/apt-get", []string{"install", "-y", "-q", "dnsmasq"}, false, ""},
-	{"dnsmasq-write", "dnsmasq drop-in write", "/usr/bin/install", []string{"-m", "0644", "/tmp/felhom-resolver-x.conf", "/etc/dnsmasq.d/felhom-x.conf"}, false, ""},
+	{"dnsmasq-write", "dnsmasq drop-in write (root content check, R-861)", "/usr/local/sbin/felhom-priv-apply", []string{"dnsmasq", "/tmp/felhom-resolver-123456789.conf", "felhom-x.conf"}, false, ""},
 	{"dnsmasq-enable", "dnsmasq enable", "/usr/bin/systemctl", []string{"enable", "--now", "dnsmasq"}, false, ""},
 
 	// ---- OS updates, guest fast lane (`11` §5.4.1; the wrapper holds every rule) ----
@@ -154,13 +151,17 @@ var manifest = []Capability{
 	// reboot-during-backup lock can't start → the customer box stays DOWN until this clears it) ----
 	{"stalelock-unlock", "reboot-during-backup stale-lock recovery", "/usr/sbin/pct", []string{"unlock", "9201"}, true, ""},
 
+	// ---- Weekly guest disk trim (FELHOM_FSTRIM, R-444). NON-critical: a missing grant means the thin pool is not
+	// reclaimed this week (the trim job WARNs per guest and the report shows the failure), not a serving outage. ----
+	{"guest-fstrim", "weekly guest disk trim (thin-pool reclaim, R-444)", "/usr/sbin/pct", []string{"fstrim", "9201"}, false, ""},
+
 	// ---- Offsite WG tunnel (FELHOM_WG, S3/v0.64.0; Critical FLIPPED in S4/v0.66.0 — offsite
 	// backups now RIDE the tunnel, so a degraded tunnel capability is operator-alert-worthy: the
 	// conf install, unit enable/restart and the handshake read gate the backup path. apt-install
 	// (one-time bootstrap) and disable (revocation, a deliberate teardown) stay non-critical. The
 	// handshake read is the ONLY wg invocation (never `dump`). ----
 	{"wg-tools-install", "wireguard-tools package install", "/usr/bin/apt-get", []string{"install", "-y", "-q", "wireguard-tools"}, false, ""},
-	{"wg-conf-install", "wg-felhom conf install", "/usr/bin/install", []string{"-o", "root", "-g", "root", "-m", "0600", "--", "/var/lib/felhom-agent/wg/wg-felhom.conf", "/etc/wireguard/wg-felhom.conf"}, true, ""},
+	{"wg-conf-install", "wg-felhom conf install (root content check, R-861)", "/usr/local/sbin/felhom-priv-apply", []string{"wg"}, true, ""},
 	{"wg-enable", "wg-quick@wg-felhom enable", "/usr/bin/systemctl", []string{"enable", "--now", "wg-quick@wg-felhom"}, true, ""},
 	{"wg-restart", "wg-quick@wg-felhom restart (conf change)", "/usr/bin/systemctl", []string{"restart", "wg-quick@wg-felhom"}, true, ""},
 	{"wg-disable", "wg-quick@wg-felhom disable (revocation)", "/usr/bin/systemctl", []string{"disable", "--now", "wg-quick@wg-felhom"}, false, ""},
@@ -192,7 +193,6 @@ var manifest = []Capability{
 	// operator-driven op, not a steady-state serving path — a degraded grant means "can't
 	// self-update" (fall back to a manual SSH deploy), not a serving outage. The apply repr uses a
 	// staging-dir path + a placeholder sha (list-mode never runs it). ----
-	{"selfupdate-apply", "agent self-update apply (A/B flip)", "/usr/local/sbin/felhom-selfupdate-guarded", []string{"apply", "/var/lib/felhom-agent/selfupdate/felhom-agent-0.0.0", "0000000000000000000000000000000000000000000000000000000000000000"}, false, ""},
 	{"selfupdate-commit", "agent self-update commit", "/usr/local/sbin/felhom-selfupdate-guarded", []string{"commit"}, false, ""},
 	{"selfupdate-rollback", "agent self-update rollback", "/usr/local/sbin/felhom-selfupdate-guarded", []string{"rollback"}, false, ""},
 }

internal/capability/manifest_test.go

@@ -56,8 +56,10 @@ func parseSudoersEntries(t *testing.T, text string) []string {
 	var cur strings.Builder
 	for i := 0; i < len(raw); i++ {
 		c := raw[i]
-		if c == '\\' && i+1 < len(raw) {
-			cur.WriteByte(raw[i+1]) // unescape: keep the next char literally (\, → , ; \: → :)
+		if c == '\\' && i+1 < len(raw) && strings.IndexByte(",:=", raw[i+1]) >= 0 {
+			// unescape the sudoers grammar escapes only (\, → , ; \: → : ; \= → =). Every other backslash is kept: in
+			// a regex entry (R-861) `\.` `\{` `\\` mean exactly what sudo's regex engine reads.
+			cur.WriteByte(raw[i+1])
 			i++
 			continue
 		}
@@ -108,10 +110,24 @@ func globToRegex(pat string) *regexp.Regexp {
 	return regexp.MustCompile(b.String())
 }
 
+// entryRegex compiles one sudoers entry. R-861 (v0.146.0): an entry whose ARGUMENTS are a sudo regular expression
+// (`^...$`, sudo >= 1.9.10) is matched as one — the binary literally, then a space, then the arguments joined by spaces
+// (sudo's own rule). Every other entry is an fnmatch glob (globToRegex). The parser has already undone the sudoers
+// escapes (`\,` `\:` `\=` `\\`), which leaves a valid RE2 pattern.
+func entryRegex(e string) *regexp.Regexp {
+	if i := strings.IndexByte(e, ' '); i > 0 {
+		bin, args := e[:i], e[i+1:]
+		if strings.HasPrefix(args, "^") && strings.HasSuffix(args, "$") {
+			return regexp.MustCompile("^" + regexp.QuoteMeta(bin) + " " + args[1:])
+		}
+	}
+	return globToRegex(e)
+}
+
 // matchesAny reports whether cmdline matches at least one sudoers entry pattern.
 func matchesAny(cmdline string, entries []string) bool {
 	for _, e := range entries {
-		if globToRegex(e).MatchString(cmdline) {
+		if entryRegex(e).MatchString(cmdline) {
 			return true
 		}
 	}

internal/capability/r861_injection_test.go

@@ -0,0 +1,95 @@
+package capability
+
+import (
+	"os"
+	"testing"
+)
+
+// R-861 (agent v0.146.0): the attacks the old globs let through, each as the exact argv a compromised agent would
+// send. NONE may match any entry of the new sudoers. The same list ran against the REAL sudo 1.9.16 (a throwaway
+// container, and `sudo -l -U felhom-agent` on both demo boxes after the bundle): audits/hub-safety-2026-10-05/partF/.
+//
+// RED-PROOF: run this list against the v0.145.0 sudoers (git show v0.145.0:configs/felhom-agent.sudoers) → most of
+// these MATCH (recorded in the same evidence folder).
+var r861Injections = []string{
+	// a raw host disk for a guest (pct options smuggled through a vmid glob)
+	"/usr/sbin/pct set 9201 --dev0 /dev/sda -onboot 1",
+	"/usr/sbin/pct set 9201 --dev0 /dev/sda -mp8 /mnt/felhom-drives",
+	"/usr/sbin/pct set 9201 --delete mp0 --dev0 /dev/sda",
+	"/usr/sbin/pct set 9201 -mp0 /var/lib/felhom-agent/guests/9201/bootstrap,mp=/x --dev0 /dev/sda",
+	// a bind mount over /etc through traversal
+	"/usr/bin/mount --bind /mnt/../var/lib/felhom-agent/x/felhom-data /mnt/felhom-drives/x",
+	"/usr/bin/mount --bind /mnt/a/felhom-data /mnt/felhom-drives/../../etc/sudoers.d",
+	"/usr/bin/umount /mnt/felhom-drives/x /",
+	"/usr/bin/chown 100000:100000 /mnt/a/felhom-data /etc/shadow",
+	"/usr/bin/mkdir -p /mnt/felhom-drives/x /etc/systemd/system/evil.mount",
+	// root-read files the agent writes: gone as `install` lines
+	"/usr/bin/install -o root -g root -m 0644 -- /var/lib/felhom-agent/units/x.mount /etc/systemd/system/etc-sudoers.d.mount",
+	"/usr/bin/install -m 0755 -- /tmp/felhom-guest-hook-1.sh /var/lib/vz/snippets/felhom-guest-hook.sh",
+	"/usr/bin/install -m 0755 -- /tmp/felhom-shared-parent-1.sh /usr/local/sbin/felhom-shared-parent.sh",
+	"/usr/bin/install -m 0644 /tmp/felhom-resolver-1.conf /etc/dnsmasq.d/felhom-x.conf",
+	"/usr/bin/install -o root -g root -m 0600 -- /var/lib/felhom-agent/wg/wg-felhom.conf /etc/wireguard/wg-felhom.conf",
+	"/usr/bin/install -o root -g root -m 0644 -- /var/lib/felhom-agent/felhom-sshd/sshd_config /etc/felhom-sshd/sshd_config",
+	// the unsigned binary flip
+	"/usr/local/sbin/felhom-selfupdate-guarded apply /var/lib/felhom-agent/selfupdate/felhom-agent-9.9.9 0000000000000000000000000000000000000000000000000000000000000000",
+	// enabling or removing anything that is not ours
+	"/usr/bin/systemctl enable --now -- mnt-hdd_1.mount evil.service",
+	"/usr/bin/systemctl enable --now -- etc-sudoers.d.mount",
+	"/usr/bin/rm -f /etc/systemd/system/mnt-felhomx /etc/passwd",
+	"/usr/bin/rm -f /etc/dnsmasq.d/felhom-x.conf /etc/shadow",
+	"/usr/bin/rmdir /mnt/felhom-drives/x /etc",
+	// nftables commands chained after a set element
+	"/usr/sbin/nft add element inet felhom_oob operator_ips { 10.77.0.250 } ; flush ruleset",
+	// extra options to read-only tools
+	"/usr/sbin/smartctl -a -j /dev/sda -s off",
+	"/usr/sbin/lvs --reportformat json --units b -o lv_name,data_percent,metadata_percent -- pve/data --config x",
+	"/usr/sbin/pct exec 9201 --keep-env -- docker inspect -f x felhom-controller",
+	"/usr/sbin/pct unlock 9201 --whatever",
+	// the checker with a path it must never take
+	"/usr/local/sbin/felhom-priv-apply unit ../../etc/x.mount",
+	"/usr/local/sbin/felhom-priv-apply dnsmasq /etc/shadow felhom-x.conf",
+	"/usr/local/sbin/felhom-priv-apply wg /etc/shadow",
+}
+
+func TestSudoersRefusesTheR861Injections(t *testing.T) {
+	data, err := os.ReadFile(sudoersPath)
+	if err != nil {
+		t.Fatal(err)
+	}
+	entries := parseSudoersEntries(t, string(data))
+	for _, c := range r861Injections {
+		if matchesAny(c, entries) {
+			t.Errorf("the sudoers still allows: %s", c)
+		}
+	}
+}
+
+// R-444: the weekly trim's grant is ONE exact shape — `pct fstrim <vmid>` — and nothing smuggled after it.
+// The manifest entry (guest-fstrim) proves the real call is still allowed (TestManifestCoveredBySudoers); this
+// pins the other direction. RED-PROOF: write the rule as the glob `/usr/sbin/pct fstrim [0-9]*` → every decoy
+// below with a trailing argument matches (the glob's `*` eats spaces).
+func TestSudoersFstrimRuleIsExact(t *testing.T) {
+	data, err := os.ReadFile(sudoersPath)
+	if err != nil {
+		t.Fatal(err)
+	}
+	entries := parseSudoersEntries(t, string(data))
+	if !matchesAny("/usr/sbin/pct fstrim 9201", entries) {
+		t.Fatal("the sudoers does not allow `pct fstrim 9201` — the weekly trim cannot run")
+	}
+	for _, c := range []string{
+		"/usr/sbin/pct fstrim 9201 --ignore-mountpoints",
+		"/usr/sbin/pct fstrim 9201 --ignore-mountpoints 1",
+		"/usr/sbin/pct fstrim 9201; x",
+		"/usr/sbin/pct fstrim 9201 9202",
+		"/usr/sbin/pct fstrim 92a1",
+		"/usr/sbin/pct fstrim ",
+		"/usr/sbin/pct fstrim -- 9201",
+		"/usr/sbin/pct destroy 9201",
+		"/usr/sbin/pct destroy 9201 --purge",
+	} {
+		if matchesAny(c, entries) {
+			t.Errorf("the sudoers allows a command the trim rule must not: %q", c)
+		}
+	}
+}

internal/config/config.go

@@ -33,6 +33,7 @@ type Config struct {
 	LANResolver LANResolverConfig `json:"lan_resolver"`
 	WGTunnel    WGTunnelConfig    `json:"wg_tunnel"`
 	GuestNet    GuestNetConfig    `json:"guest_net"`
+	DiskTrim    DiskTrimConfig    `json:"disk_trim"`
 	OOB         OOBConfig         `json:"oob"`
 	SelfUpdate  SelfUpdateConfig  `json:"selfupdate"`
 	LogLevel    string            `json:"log_level"` // debug|info|warn|error (default info)
@@ -139,6 +140,16 @@ func (w WGTunnelConfig) WithDefaults() WGTunnelConfig {
 	return w
 }
 
+// DiskTrimConfig configures the R-444 weekly guest disk trim (internal/fstrim). DEFAULT-ON, like GuestNetConfig and for
+// the same reason: it only acts on guests the agent already owns, and the operator ruled every box trims (`09` §3
+// decision 139). Opting out is the explicit act: `"disk_trim": {"disable": true}`.
+type DiskTrimConfig struct {
+	Disable bool `json:"disable"`
+}
+
+// Enabled reports whether the weekly trim should run.
+func (d DiskTrimConfig) Enabled() bool { return !d.Disable }
+
 // GuestNetConfig configures the R-54 guest-network watchdog (internal/guestnet).
 //
 // **This is the repo's first DEFAULT-ON feature gate, and the inversion is deliberate.** Every other

internal/escrow/identity.go

@@ -7,9 +7,11 @@ import (
 	"encoding/hex"
 	"encoding/json"
 	"fmt"
+	"io"
 	"os"
 	"path/filepath"
 	"strings"
+	"syscall"
 )
 
 // Slice 10D.1 — IDENTITY escrow. The K-escrow (above) wraps the PBS *encryption key* via the
@@ -71,7 +73,7 @@ func HashResticPassword(pw string) string {
 // (non-empty) but NEVER logged by callers — log the field NAME only (mirrors AttachWGKey). A missing file
 // is a clean no-attach (pre-fork-4 behavior, byte-compatible bundle).
 func AttachResticPassword(b *IdentityBundle, stagePath string) (bool, error) {
-	raw, err := os.ReadFile(stagePath)
+	raw, err := readStagedNoFollow(stagePath)
 	if err != nil {
 		if os.IsNotExist(err) {
 			return false, nil
@@ -92,7 +94,7 @@ func AttachResticPassword(b *IdentityBundle, stagePath string) (bool, error) {
 // no-attach (pre-S3 behavior, byte-compatible bundle); a corrupt one is an error (the operator
 // should know their escrow would silently lack a live identity).
 func AttachWGKey(b *IdentityBundle, keyPath string) (bool, error) {
-	raw, err := os.ReadFile(keyPath)
+	raw, err := readStagedNoFollow(keyPath)
 	if err != nil {
 		if os.IsNotExist(err) {
 			return false, nil
@@ -180,3 +182,47 @@ func UnwrapIdentityBundle(ctx context.Context, blob []byte, recoveryCode string)
 	}
 	return b, nil
 }
+
+// readStagedNoFollow reads a file the AGENT staged, for the escrow ceremony that runs as ROOT (FELHOM_ESCROW). R-861
+// (agent v0.146.0/0.146.1): both files live in the agent's own directory, so a compromised agent could put a SYMLINK
+// there — at the file OR at any directory on the way (review 2026-10-05) — to a root-only file, and the root ceremony
+// would seal that file into the blob and hand the agent R. So the path is walked from "/" one component at a time with
+// openat(O_NOFOLLOW): no symlink anywhere, the last a regular file of at most 4 KiB. Once a directory is open, renaming
+// it does not redirect the walk. A missing file keeps its os.IsNotExist meaning. Pinned by TestAttach_RefusesASymlink*.
+func readStagedNoFollow(path string) ([]byte, error) {
+	if !filepath.IsAbs(path) {
+		return nil, fmt.Errorf("%s is not an absolute path", path)
+	}
+	clean := filepath.Clean(path)
+	parts := strings.Split(strings.TrimPrefix(clean, "/"), "/")
+	dirfd, err := syscall.Open("/", syscall.O_RDONLY|syscall.O_DIRECTORY|syscall.O_CLOEXEC, 0)
+	if err != nil {
+		return nil, err
+	}
+	for i, part := range parts {
+		last := i == len(parts)-1
+		flags := syscall.O_RDONLY | syscall.O_NOFOLLOW | syscall.O_CLOEXEC
+		if !last {
+			flags |= syscall.O_DIRECTORY
+		}
+		fd, err := syscall.Openat(dirfd, part, flags, 0)
+		syscall.Close(dirfd)
+		if err != nil {
+			return nil, &os.PathError{Op: "open", Path: clean, Err: err}
+		}
+		dirfd = fd
+	}
+	f := os.NewFile(uintptr(dirfd), clean)
+	defer f.Close()
+	fi, err := f.Stat()
+	if err != nil {
+		return nil, err
+	}
+	if !fi.Mode().IsRegular() {
+		return nil, fmt.Errorf("%s is not a regular file", path)
+	}
+	if fi.Size() > 4096 {
+		return nil, fmt.Errorf("%s is larger than 4 KiB", path)
+	}
+	return io.ReadAll(io.LimitReader(f, 4097))
+}

internal/escrow/r861_staged_read_test.go

@@ -0,0 +1,63 @@
+package escrow
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+// R-861 (agent v0.146.0): the root escrow ceremony reads two files from the AGENT's directory. A symlink there to a
+// root-only file must never be read (it would be sealed under R and handed to the agent).
+// RED-PROOF (audits/hub-safety-2026-10-05/partF/red-proof.txt): use os.ReadFile in readStagedNoFollow → this fails.
+func TestAttach_RefusesASymlink(t *testing.T) {
+	d := t.TempDir()
+	secret := filepath.Join(d, "root-only")
+	if err := os.WriteFile(secret, []byte("ROOT-ONLY-CANARY"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	link := filepath.Join(d, "restic_repo_password")
+	if err := os.Symlink(secret, link); err != nil {
+		t.Fatal(err)
+	}
+	var b IdentityBundle
+	if ok, err := AttachResticPassword(&b, link); err == nil || ok || b.ResticRepoPassword != "" {
+		t.Fatalf("a symlinked staged file was read: ok=%v err=%v value-set=%v", ok, err, b.ResticRepoPassword != "")
+	}
+	if ok, err := AttachWGKey(&b, link); err == nil || ok {
+		t.Fatalf("a symlinked wg key was read: ok=%v err=%v", ok, err)
+	}
+	// control: the real staged file is still read; a missing one is still a clean no-attach
+	real := filepath.Join(d, "real")
+	_ = os.WriteFile(real, []byte("pw\n"), 0o600)
+	if ok, err := AttachResticPassword(&b, real); err != nil || !ok || b.ResticRepoPassword != "pw" {
+		t.Fatalf("control: the plain staged file was not read: %v %v", ok, err)
+	}
+	if ok, err := AttachResticPassword(&b, filepath.Join(d, "absent")); err != nil || ok {
+		t.Fatalf("control: a missing file must stay a clean no-attach: %v %v", ok, err)
+	}
+}
+
+// Review 2026-10-05: a symlinked DIRECTORY on the way must stop the read too (O_NOFOLLOW alone guards only the last
+// component). RED-PROOF: open the full path with O_NOFOLLOW only → this fails.
+func TestAttach_RefusesASymlinkedDirectory(t *testing.T) {
+	d := t.TempDir()
+	secretDir := filepath.Join(d, "root-only-dir")
+	_ = os.Mkdir(secretDir, 0o700)
+	_ = os.WriteFile(filepath.Join(secretDir, "private.key"), []byte("AAECAwQFBgcICQoLDA0ODxAREhMUFRYXGBkaGxwdHh8=\n"), 0o600)
+	agentDir := filepath.Join(d, "agent")
+	_ = os.Mkdir(agentDir, 0o700)
+	if err := os.Symlink(secretDir, filepath.Join(agentDir, "wg")); err != nil {
+		t.Fatal(err)
+	}
+	var b IdentityBundle
+	if ok, err := AttachWGKey(&b, filepath.Join(agentDir, "wg", "private.key")); err == nil || ok || b.WGPrivateKey != "" {
+		t.Fatalf("a key behind a symlinked directory was read: ok=%v err=%v", ok, err)
+	}
+	// control: the same key under a REAL directory is read
+	_ = os.Remove(filepath.Join(agentDir, "wg"))
+	_ = os.Mkdir(filepath.Join(agentDir, "wg"), 0o700)
+	_ = os.WriteFile(filepath.Join(agentDir, "wg", "private.key"), []byte("AAECAwQFBgcICQoLDA0ODxAREhMUFRYXGBkaGxwdHh8=\n"), 0o600)
+	if ok, err := AttachWGKey(&b, filepath.Join(agentDir, "wg", "private.key")); err != nil || !ok {
+		t.Fatalf("control: a key under a real directory was not read: %v %v", ok, err)
+	}
+}

internal/felhomsshd/config.go

@@ -27,6 +27,8 @@ const (
 	PortFile = ConfDir + "/port"
 	// PidFile is the instance pidfile (NOT a RuntimeDirectory — that is the G1 incident cause).
 	PidFile = "/run/felhom-sshd.pid"
+	// PrivApply is the root content checker that installs the config and felhom-op's key (R-861, agent v0.146.0).
+	PrivApply = "/usr/local/sbin/felhom-priv-apply"
 	// Unit is the systemd unit name.
 	Unit = "felhom-sshd"
 	// OperatorUser is the default operator login (scoped sudo; key in AuthKeysDir only).

internal/felhomsshd/manager.go

@@ -143,7 +143,8 @@ func (m *Manager) Apply(ctx context.Context, block *hub.WireWireguard) (int, err
 		m.logger.Error("felhomsshd: staged config failed sshd -t — NOT installing", "err", err, "stderr", strings.TrimSpace(string(errOut)))
 		return port, err
 	}
-	if _, errOut, err := m.runner.Run(ctx, "install", "-o", "root", "-g", "root", "-m", "0644", "--", m.stagedConfPath(), ConfPath); err != nil {
+	// R-861 (v0.146.0): the root checker installs it, and only if it is renderConfig's template for some port.
+	if _, errOut, err := m.runner.Run(ctx, PrivApply, "sshd-config"); err != nil {
 		m.logger.Error("felhomsshd: config install failed", "err", err, "stderr", strings.TrimSpace(string(errOut)))
 		return port, err
 	}
@@ -181,7 +182,9 @@ func (m *Manager) applyAuthorizedKeys(ctx context.Context, sshKey string) {
 		m.logger.Error("felhomsshd: staging authorized_keys", "err", err)
 		return
 	}
-	if _, errOut, err := m.runner.Run(ctx, "install", "-o", "root", "-g", "root", "-m", "0644", "--", staged, AuthKeysUserPath); err != nil {
+	// R-861: the root checker installs it — one plain public key, no options (command=, from=, …), or empty.
+	_ = staged
+	if _, errOut, err := m.runner.Run(ctx, PrivApply, "sshd-key"); err != nil {
 		m.logger.Error("felhomsshd: authorized_keys install failed", "err", err, "stderr", strings.TrimSpace(string(errOut)))
 		return
 	}

internal/felhomsshd/r861_privapply_contract_test.go

@@ -0,0 +1,26 @@
+package felhomsshd
+
+import (
+	"strings"
+	"testing"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/privapplytest"
+)
+
+// R-861: renderConfig is byte-identical to the checker's template (only the Port varies); a changed line is refused.
+// RED-PROOF: change one directive in renderConfig → this fails (the two templates drifted).
+func TestPrivApply_AcceptsTheRenderedConfig(t *testing.T) {
+	for _, port := range []int{2222, 8822, 60022} {
+		conf, err := renderConfig(port)
+		if err != nil {
+			t.Fatal(err)
+		}
+		if got := privapplytest.Check(t, "sshd-config", "", conf); got != "OK" {
+			t.Errorf("port %d: %s", port, got)
+		}
+	}
+	conf, _ := renderConfig(8822)
+	if got := privapplytest.Check(t, "sshd-config", "", conf+"StrictModes no\n"); !strings.HasPrefix(got, "REFUSED") {
+		t.Fatalf("control: an extra directive was not refused: %s", got)
+	}
+}

internal/fstrim/fstrim.go

@@ -0,0 +1,346 @@
+// Package fstrim is the weekly guest disk trim (R-444, operator ruling `09` §3 decision 139).
+//
+// Why: a thin pool only ever grows from blocks a guest has already FREED — `fstrim` inside the unprivileged container
+// is refused (FITRIM: Operation not permitted), and nothing else on the box gives the blocks back. A full thin pool
+// takes every guest on the host read-only, so the pool can reach 100 % from deleted data alone. Measured on demo-hp
+// 2026-10-06 09:14Z: `pct fstrim 9201` rc 0 in 24.4 s, pool 65.53 % -> 33.40 %, 18/18 app probes 200, max 1.1 s
+// (audits/ten-answers-2026-10-06/r444-measure.txt).
+//
+// The rule, each part pinned by a test in fstrim_test.go:
+//   - Weekly: a guest is DUE from Wednesday 10:00 local until it has been trimmed once since then (a box that was off
+//     on Wednesday catches up at its next eligible hour).
+//   - Daytime only: a trim starts only between 10:00 and 20:59 local — never in the night window (01:00–06:59) where
+//     the backups and the restore-tests run (TestEligibleHourNeverInTheNight).
+//   - Never beside a backup, a restore-test or another heavy operation: the pass holds the host-wide one-heavy-op gate
+//     (backup.InFlight) for its whole run; a busy gate DEFERS the pass to the next hourly tick.
+//   - A failed trim is retried at the next eligible hour, at most MaxAttemptsPerWeek times in one week.
+//   - The last result per guest (time, bytes, ok/fail) is persisted, so a restart neither loses it nor re-trims.
+//
+// The command is the ONE exact sudoers shape `pct fstrim <vmid>` (FELHOM_FSTRIM). Only guests from the pool-verified
+// source (ListLXC ∩ the felhom pool, audit A1) and only RUNNING ones are trimmed.
+package fstrim
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"log/slog"
+	"os"
+	"path/filepath"
+	"regexp"
+	"sort"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/hub"
+	"gitea.dooplex.hu/admin/felhom-agent/internal/proxmox"
+)
+
+// Schedule. The weekday/hours are fixed on purpose (one sentence the operator can read on the System page).
+const (
+	Weekday            = time.Wednesday
+	StartHour          = 10 // first eligible local hour (inclusive)
+	EndHour            = 21 // first NOT-eligible local hour (exclusive): last start is 20:59
+	MaxAttemptsPerWeek = 3
+	// TickInterval is how often the job looks; a deferred or failed pass is therefore retried the next hour.
+	TickInterval = time.Hour
+	// FirstTickDelay lets the agent settle after a start before the first look.
+	FirstTickDelay = 5 * time.Minute
+	// PerGuestTimeout bounds one `pct fstrim` (measured 24.4 s for 84 GiB).
+	PerGuestTimeout = 30 * time.Minute
+)
+
+// ScheduleText is the human description carried on the host report.
+const ScheduleText = "weekly, due Wednesday from 10:00 host-local time; starts only 10:00-20:59; never beside a backup or restore-test"
+
+// Runner runs a host command (proxmox.ExecRunner in production, through `sudo -n`).
+type Runner interface {
+	Run(ctx context.Context, name string, args ...string) (stdout, stderr []byte, err error)
+}
+
+// GuestSource yields the guests this agent OWNS (the pool-verified source, never a bare ListLXC).
+type GuestSource interface {
+	Guests(ctx context.Context) ([]proxmox.Guest, error)
+}
+
+// Gate is the host-wide one-heavy-operation gate (*backup.InFlight).
+type Gate interface {
+	TryAcquire(what string) (release func(), busy string, ok bool)
+}
+
+// GateName is what the gate reports as busy while a trim runs.
+const GateName = "guest-fstrim"
+
+// Record is one guest's last trim attempt, as persisted.
+type Record struct {
+	LastAttemptAt   time.Time `json:"last_attempt_at"`
+	OK              bool      `json:"ok"`
+	BytesTrimmed    int64     `json:"bytes_trimmed"`
+	Mounts          int       `json:"mounts"`
+	DurationSeconds float64   `json:"duration_seconds"`
+	LastOKAt        time.Time `json:"last_ok_at,omitempty"`
+	Error           string    `json:"error,omitempty"`
+	// Attempts counts the attempts since the current week's due time (reset by the first attempt of a new week).
+	Attempts int `json:"attempts"`
+}
+
+// Trimmer is the weekly job.
+type Trimmer struct {
+	runner    Runner
+	guests    GuestSource
+	gate      Gate
+	statePath string
+	logger    *slog.Logger
+	loc       *time.Location
+	now       func() time.Time
+
+	mu      sync.Mutex
+	records map[int]Record
+}
+
+// New builds the job and loads the persisted state. A missing state file is an empty state; a corrupt one is logged
+// and treated as empty (the cost is one extra trim, never a missed one).
+func New(runner Runner, guests GuestSource, gate Gate, statePath string, logger *slog.Logger) *Trimmer {
+	if logger == nil {
+		logger = slog.Default()
+	}
+	t := &Trimmer{runner: runner, guests: guests, gate: gate, statePath: statePath, logger: logger,
+		loc: time.Local, now: time.Now, records: map[int]Record{}}
+	t.load()
+	return t
+}
+
+func (t *Trimmer) load() {
+	data, err := os.ReadFile(t.statePath)
+	if err != nil {
+		if !os.IsNotExist(err) {
+			t.logger.Warn("fstrim: state read failed — starting empty", "path", t.statePath, "err", err)
+		}
+		return
+	}
+	var raw map[string]Record
+	if err := json.Unmarshal(data, &raw); err != nil {
+		t.logger.Warn("fstrim: state file corrupt — starting empty", "path", t.statePath, "err", err)
+		return
+	}
+	for k, r := range raw {
+		if id, err := strconv.Atoi(k); err == nil && id > 0 {
+			t.records[id] = r
+		}
+	}
+}
+
+func (t *Trimmer) saveLocked() error {
+	raw := make(map[string]Record, len(t.records))
+	for id, r := range t.records {
+		raw[strconv.Itoa(id)] = r
+	}
+	data, err := json.MarshalIndent(raw, "", "  ")
+	if err != nil {
+		return err
+	}
+	if err := os.MkdirAll(filepath.Dir(t.statePath), 0o755); err != nil {
+		return err
+	}
+	tmp := t.statePath + ".tmp"
+	if err := os.WriteFile(tmp, data, 0o600); err != nil {
+		os.Remove(tmp)
+		return err
+	}
+	return os.Rename(tmp, t.statePath)
+}
+
+// EligibleHour reports whether a trim may START at local time lt.
+func EligibleHour(lt time.Time) bool {
+	h := lt.Hour()
+	return h >= StartHour && h < EndHour
+}
+
+// weekAnchor is the most recent Wednesday StartHour:00 at or before lt (same location as lt).
+func weekAnchor(lt time.Time) time.Time {
+	daysBack := (int(lt.Weekday()) - int(Weekday) + 7) % 7
+	d := lt.AddDate(0, 0, -daysBack)
+	a := time.Date(d.Year(), d.Month(), d.Day(), StartHour, 0, 0, 0, lt.Location())
+	if a.After(lt) {
+		d = d.AddDate(0, 0, -7)
+		a = time.Date(d.Year(), d.Month(), d.Day(), StartHour, 0, 0, 0, lt.Location())
+	}
+	return a
+}
+
+// due reports whether a guest with record r (ok=false: none) is due at local time lt.
+func due(r Record, has bool, lt time.Time) bool {
+	if !has {
+		return true
+	}
+	anchor := weekAnchor(lt)
+	if r.LastAttemptAt.Before(anchor) {
+		return true // not tried this week
+	}
+	return !r.OK && r.Attempts < MaxAttemptsPerWeek
+}
+
+// Run looks every TickInterval until ctx ends. It never returns an error: a failed trim is a reported fact.
+func (t *Trimmer) Run(ctx context.Context) {
+	t.logger.Info("fstrim: weekly guest disk trim starting", "schedule", ScheduleText)
+	timer := time.NewTimer(FirstTickDelay)
+	defer timer.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-timer.C:
+			t.Pass(ctx)
+			timer.Reset(TickInterval)
+		}
+	}
+}
+
+// Pass is one look: outside the daytime window it does nothing; otherwise it trims every due, running, owned guest
+// while holding the heavy-op gate.
+func (t *Trimmer) Pass(ctx context.Context) {
+	lt := t.now().In(t.loc)
+	if !EligibleHour(lt) {
+		t.logger.Debug("fstrim: outside the daytime window — not looking", "local", lt.Format("Mon 15:04"))
+		return
+	}
+	guests, err := t.guests.Guests(ctx)
+	if err != nil {
+		t.logger.Warn("fstrim: owned-guest list unavailable — skipping this pass", "err", err)
+		return
+	}
+	owned := make(map[int]bool, len(guests))
+	var todo []int
+	t.mu.Lock()
+	for _, g := range guests {
+		owned[g.VMID] = true
+		r, has := t.records[g.VMID]
+		if !due(r, has, lt) {
+			continue
+		}
+		if g.Status != "running" {
+			t.logger.Info("fstrim: guest not running — trimmed when it runs", "vmid", g.VMID, "status", g.Status)
+			continue
+		}
+		todo = append(todo, g.VMID)
+	}
+	// A guest the agent no longer owns has no result to report.
+	pruned := false
+	for id := range t.records {
+		if !owned[id] {
+			delete(t.records, id)
+			pruned = true
+		}
+	}
+	if pruned {
+		if err := t.saveLocked(); err != nil {
+			t.logger.Warn("fstrim: state save failed", "err", err)
+		}
+	}
+	t.mu.Unlock()
+	if len(todo) == 0 {
+		return
+	}
+	sort.Ints(todo)
+	release, busy, ok := t.gate.TryAcquire(GateName)
+	if !ok {
+		t.logger.Info("fstrim: deferred — a heavy operation is in flight; retrying next hour", "busy", busy, "due_guests", len(todo))
+		return
+	}
+	defer release()
+	for _, vmid := range todo {
+		if ctx.Err() != nil {
+			return
+		}
+		t.trimOne(ctx, vmid, lt)
+	}
+}
+
+var trimmedLine = regexp.MustCompile(`\((\d+) bytes\) trimmed`)
+
+// ParseTrimmed sums the "(N bytes) trimmed" lines of `pct fstrim` output and counts them (one per mount point), e.g.
+// `/var/lib/lxc/9201/rootfs/: 30.1 GiB (32277680128 bytes) trimmed`.
+func ParseTrimmed(out string) (bytes int64, mounts int) {
+	for _, m := range trimmedLine.FindAllStringSubmatch(out, -1) {
+		n, err := strconv.ParseInt(m[1], 10, 64)
+		if err != nil {
+			continue
+		}
+		bytes += n
+		mounts++
+	}
+	return bytes, mounts
+}
+
+// GiB renders bytes as "30.1 GiB".
+func GiB(b int64) string { return fmt.Sprintf("%.1f GiB", float64(b)/(1<<30)) }
+
+func (t *Trimmer) trimOne(ctx context.Context, vmid int, lt time.Time) {
+	start := t.now()
+	cctx, cancel := context.WithTimeout(ctx, PerGuestTimeout)
+	stdout, stderr, err := t.runner.Run(cctx, "pct", "fstrim", strconv.Itoa(vmid))
+	cancel()
+	dur := t.now().Sub(start)
+	bytes, mounts := ParseTrimmed(string(stdout) + "\n" + string(stderr))
+
+	t.mu.Lock()
+	prev, has := t.records[vmid]
+	r := Record{LastAttemptAt: start.UTC(), OK: err == nil, BytesTrimmed: bytes, Mounts: mounts,
+		DurationSeconds: float64(dur.Round(100*time.Millisecond)) / float64(time.Second), LastOKAt: prev.LastOKAt}
+	if has && !prev.LastAttemptAt.Before(weekAnchor(lt)) {
+		r.Attempts = prev.Attempts + 1
+	} else {
+		r.Attempts = 1
+	}
+	if err == nil {
+		r.LastOKAt = start.UTC()
+	} else {
+		msg := strings.TrimSpace(err.Error() + ": " + strings.TrimSpace(string(stderr)))
+		if len(msg) > 300 {
+			msg = msg[:300]
+		}
+		r.Error = msg
+	}
+	t.records[vmid] = r
+	saveErr := t.saveLocked()
+	t.mu.Unlock()
+
+	if err == nil {
+		t.logger.Info(fmt.Sprintf("fstrim: guest %d trimmed %s in %.1fs", vmid, GiB(bytes), r.DurationSeconds),
+			"vmid", vmid, "bytes_trimmed", bytes, "mounts", mounts, "duration_s", r.DurationSeconds)
+		if mounts == 0 {
+			t.logger.Warn("fstrim: pct fstrim succeeded but reported no trimmed mount — output not understood",
+				"vmid", vmid, "stdout", strings.TrimSpace(string(stdout)))
+		}
+	} else {
+		t.logger.Warn(fmt.Sprintf("fstrim: guest %d trim FAILED after %.1fs", vmid, r.DurationSeconds),
+			"vmid", vmid, "attempt", r.Attempts, "max_attempts_per_week", MaxAttemptsPerWeek, "err", r.Error)
+	}
+	if saveErr != nil {
+		t.logger.Warn("fstrim: state save failed — the result will not survive a restart", "path", t.statePath, "err", saveErr)
+	}
+}
+
+// GuestDiskTrimStatus implements hub.GuestDiskTrimReporter: a pure read of the persisted results (never runs pct).
+func (t *Trimmer) GuestDiskTrimStatus(context.Context) *hub.GuestDiskTrimStatus {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+	out := &hub.GuestDiskTrimStatus{Schedule: ScheduleText}
+	ids := make([]int, 0, len(t.records))
+	for id := range t.records {
+		ids = append(ids, id)
+	}
+	sort.Ints(ids)
+	for _, id := range ids {
+		r := t.records[id]
+		g := hub.GuestDiskTrim{VMID: id, LastAttemptAt: r.LastAttemptAt.UTC().Format(time.RFC3339), OK: r.OK,
+			BytesTrimmed: r.BytesTrimmed, Mounts: r.Mounts, DurationSeconds: r.DurationSeconds, Error: r.Error}
+		if !r.LastOKAt.IsZero() {
+			g.LastOKAt = r.LastOKAt.UTC().Format(time.RFC3339)
+		}
+		out.Guests = append(out.Guests, g)
+	}
+	return out
+}

internal/fstrim/fstrim_test.go

@@ -0,0 +1,267 @@
+package fstrim
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"errors"
+	"log/slog"
+	"path/filepath"
+	"reflect"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/backup"
+	"gitea.dooplex.hu/admin/felhom-agent/internal/proxmox"
+)
+
+// The real `pct fstrim 9201` output measured on demo-hp 2026-10-06 (audits/ten-answers-2026-10-06/r444-measure.txt).
+const measuredOut = "/var/lib/lxc/9201/rootfs/: 30.1 GiB (32277680128 bytes) trimmed\n" +
+	"/var/lib/lxc/9201/rootfs/var/lib/felhom: 53.9 GiB (57865633792 bytes) trimmed\n"
+
+const measuredBytes = int64(32277680128 + 57865633792)
+
+type fakeRunner struct {
+	mu    sync.Mutex
+	calls [][]string
+	out   string
+	err   error
+	onRun func()
+}
+
+func (f *fakeRunner) Run(_ context.Context, name string, args ...string) ([]byte, []byte, error) {
+	f.mu.Lock()
+	f.calls = append(f.calls, append([]string{name}, args...))
+	f.mu.Unlock()
+	if f.onRun != nil {
+		f.onRun()
+	}
+	if f.err != nil {
+		return nil, []byte("mount busy"), f.err
+	}
+	return []byte(f.out), nil, nil
+}
+
+type fakeGuests struct {
+	g   []proxmox.Guest
+	err error
+}
+
+func (f fakeGuests) Guests(context.Context) ([]proxmox.Guest, error) { return f.g, f.err }
+
+// A Wednesday 10:30 in a fixed zone (CEST-like), so the tests do not depend on the machine's zone.
+var zone = time.FixedZone("CEST", 2*3600)
+
+func at(day, hour, min int) time.Time { return time.Date(2026, 10, day, hour, min, 0, 0, zone) } // 2026-10-07 = Wednesday
+
+func newT(t *testing.T, r Runner, g GuestSource, gate Gate, now *time.Time) (*Trimmer, *bytes.Buffer, string) {
+	t.Helper()
+	var logs bytes.Buffer
+	path := filepath.Join(t.TempDir(), "guest-disk-trim.json")
+	tr := New(r, g, gate, path, slog.New(slog.NewTextHandler(&logs, &slog.HandlerOptions{Level: slog.LevelDebug})))
+	tr.loc = zone
+	tr.now = func() time.Time { return *now }
+	return tr, &logs, path
+}
+
+func running(ids ...int) fakeGuests {
+	var g []proxmox.Guest
+	for _, id := range ids {
+		g = append(g, proxmox.Guest{VMID: id, Status: "running", Type: "lxc"})
+	}
+	return fakeGuests{g: g}
+}
+
+func TestParseTrimmedTheMeasuredOutput(t *testing.T) {
+	b, m := ParseTrimmed(measuredOut)
+	if b != measuredBytes || m != 2 {
+		t.Fatalf("ParseTrimmed = %d bytes over %d mounts, want %d over 2", b, m, measuredBytes)
+	}
+	if b, m := ParseTrimmed("something else\n"); b != 0 || m != 0 {
+		t.Fatalf("unrelated output parsed as %d/%d", b, m)
+	}
+	if got := GiB(measuredBytes); got != "84.0 GiB" {
+		t.Fatalf("GiB = %q", got)
+	}
+}
+
+// The night window (01:00–06:59) must never be eligible, and the daytime window is exactly 10:00–20:59.
+func TestEligibleHourNeverInTheNight(t *testing.T) {
+	for h := 0; h < 24; h++ {
+		lt := time.Date(2026, 10, 7, h, 30, 0, 0, zone)
+		got := EligibleHour(lt)
+		if h >= 1 && h <= 6 && got {
+			t.Errorf("hour %02d is in the night window and must not be eligible", h)
+		}
+		if want := h >= 10 && h <= 20; got != want {
+			t.Errorf("EligibleHour(%02d:30) = %v, want %v", h, got, want)
+		}
+	}
+}
+
+func TestWeekAnchorIsTheLastWednesdayTen(t *testing.T) {
+	cases := map[time.Time]time.Time{
+		at(7, 10, 0):  at(7, 10, 0),                              // Wednesday 10:00 itself
+		at(7, 9, 59):  time.Date(2026, 9, 30, 10, 0, 0, 0, zone), // before 10:00 Wednesday → the previous week
+		at(8, 15, 0):  at(7, 10, 0),                              // Thursday
+		at(13, 20, 0): at(7, 10, 0),                              // next Tuesday
+		at(14, 11, 0): at(14, 10, 0),                             // next Wednesday
+	}
+	for in, want := range cases {
+		if got := weekAnchor(in); !got.Equal(want) {
+			t.Errorf("weekAnchor(%s) = %s, want %s", in.Format("Mon 01-02 15:04"), got.Format("Mon 01-02 15:04"), want.Format("Mon 01-02 15:04"))
+		}
+	}
+}
+
+// The consequence: on Wednesday 10:30 a running owned guest is trimmed with the ONE exact argv, the bytes are parsed,
+// the positive log line is written, the result is persisted, and the host report carries it.
+func TestPassTrimsADueGuestAndReportsIt(t *testing.T) {
+	now := at(7, 10, 30)
+	r := &fakeRunner{out: measuredOut}
+	tr, logs, path := newT(t, r, running(9201), &backup.InFlight{}, &now)
+	tr.Pass(context.Background())
+
+	if want := [][]string{{"pct", "fstrim", "9201"}}; !reflect.DeepEqual(r.calls, want) {
+		t.Fatalf("runner calls = %q, want %q", r.calls, want)
+	}
+	if !strings.Contains(logs.String(), "fstrim: guest 9201 trimmed 84.0 GiB in ") {
+		t.Fatalf("no positive per-guest log line:\n%s", logs.String())
+	}
+	st := tr.GuestDiskTrimStatus(context.Background())
+	if st == nil || st.Schedule != ScheduleText || len(st.Guests) != 1 {
+		t.Fatalf("report stanza = %+v", st)
+	}
+	g := st.Guests[0]
+	if g.VMID != 9201 || !g.OK || g.BytesTrimmed != measuredBytes || g.Mounts != 2 || g.LastOKAt == "" || g.LastAttemptAt == "" {
+		t.Fatalf("report guest = %+v", g)
+	}
+	// Persisted: a NEW Trimmer over the same file (an agent restart) still has it and does not trim again this week.
+	now = at(8, 11, 0)
+	r2 := &fakeRunner{out: measuredOut}
+	tr2 := New(r2, running(9201), &backup.InFlight{}, path, slog.New(slog.NewTextHandler(&bytes.Buffer{}, nil)))
+	tr2.loc, tr2.now = zone, func() time.Time { return now }
+	if st2 := tr2.GuestDiskTrimStatus(context.Background()); len(st2.Guests) != 1 || st2.Guests[0].BytesTrimmed != measuredBytes {
+		t.Fatalf("result lost over a restart: %+v", st2)
+	}
+	tr2.Pass(context.Background())
+	if len(r2.calls) != 0 {
+		t.Fatalf("trimmed again in the same week after a restart: %q", r2.calls)
+	}
+	// Next week it is due again.
+	now = at(14, 10, 5)
+	tr2.Pass(context.Background())
+	if len(r2.calls) != 1 {
+		t.Fatalf("not trimmed in the next week: %q", r2.calls)
+	}
+}
+
+func TestPassNeverRunsInTheNight(t *testing.T) {
+	for _, h := range []int{1, 3, 6, 9, 21, 23} {
+		now := at(7, h, 15)
+		r := &fakeRunner{out: measuredOut}
+		tr, _, _ := newT(t, r, running(9201), &backup.InFlight{}, &now)
+		tr.Pass(context.Background())
+		if len(r.calls) != 0 {
+			t.Errorf("trimmed at %02d:15: %q", h, r.calls)
+		}
+	}
+}
+
+// A backup (or restore-test) holding the heavy-op gate DEFERS the trim; the next hour, gate free, it runs. And while
+// a trim runs, the gate is held, so a backup cannot start beside it.
+func TestPassDefersToAHeavyOperationAndRetriesNextHour(t *testing.T) {
+	now := at(7, 10, 30)
+	gate := &backup.InFlight{}
+	release, _, _ := gate.TryAcquire("backup:9201")
+	var busyDuringTrim string
+	r := &fakeRunner{out: measuredOut}
+	r.onRun = func() { busyDuringTrim = gate.Busy() }
+	tr, logs, _ := newT(t, r, running(9201), gate, &now)
+
+	tr.Pass(context.Background())
+	if len(r.calls) != 0 {
+		t.Fatalf("trimmed beside a running backup: %q", r.calls)
+	}
+	if !strings.Contains(logs.String(), "fstrim: deferred") || !strings.Contains(logs.String(), "backup:9201") {
+		t.Fatalf("the deferral is not logged with what holds the gate:\n%s", logs.String())
+	}
+	release()
+	now = now.Add(time.Hour)
+	tr.Pass(context.Background())
+	if len(r.calls) != 1 {
+		t.Fatalf("not retried the next hour: %q", r.calls)
+	}
+	if busyDuringTrim != GateName {
+		t.Fatalf("the heavy-op gate was %q during the trim, want %q", busyDuringTrim, GateName)
+	}
+	if gate.Busy() != "" {
+		t.Fatalf("the gate was not released after the pass: %q", gate.Busy())
+	}
+}
+
+func TestFailedTrimWarnsIsRecordedAndRetriedAtMostThreeTimes(t *testing.T) {
+	now := at(7, 10, 30)
+	r := &fakeRunner{err: errors.New("exit status 255")}
+	tr, logs, _ := newT(t, r, running(9201), &backup.InFlight{}, &now)
+	for i := 0; i < 6; i++ {
+		tr.Pass(context.Background())
+		now = now.Add(time.Hour)
+	}
+	if len(r.calls) != MaxAttemptsPerWeek {
+		t.Fatalf("attempts in one week = %d, want %d", len(r.calls), MaxAttemptsPerWeek)
+	}
+	if !strings.Contains(logs.String(), "level=WARN") || !strings.Contains(logs.String(), "fstrim: guest 9201 trim FAILED") {
+		t.Fatalf("no WARN for the failure:\n%s", logs.String())
+	}
+	g := tr.GuestDiskTrimStatus(context.Background()).Guests[0]
+	if g.OK || g.LastOKAt != "" || !strings.Contains(g.Error, "exit status 255") || !strings.Contains(g.Error, "mount busy") {
+		t.Fatalf("failed result not recorded as a failure: %+v", g)
+	}
+	// A success later keeps a clean record.
+	r.err = nil
+	r.out = measuredOut
+	now = at(14, 10, 10)
+	tr.Pass(context.Background())
+	if g := tr.GuestDiskTrimStatus(context.Background()).Guests[0]; !g.OK || g.Error != "" || g.BytesTrimmed != measuredBytes {
+		t.Fatalf("success after failure: %+v", g)
+	}
+}
+
+func TestOnlyRunningOwnedGuestsAndAFailedListActsOnNothing(t *testing.T) {
+	now := at(7, 10, 30)
+	r := &fakeRunner{out: measuredOut}
+	g := fakeGuests{g: []proxmox.Guest{{VMID: 9201, Status: "stopped"}, {VMID: 9202, Status: "running"}}}
+	tr, _, _ := newT(t, r, g, &backup.InFlight{}, &now)
+	tr.Pass(context.Background())
+	if want := [][]string{{"pct", "fstrim", "9202"}}; !reflect.DeepEqual(r.calls, want) {
+		t.Fatalf("calls = %q, want only the running guest", r.calls)
+	}
+	r2 := &fakeRunner{out: measuredOut}
+	tr2, logs, _ := newT(t, r2, fakeGuests{err: errors.New("pool read 403")}, &backup.InFlight{}, &now)
+	tr2.Pass(context.Background())
+	if len(r2.calls) != 0 || !strings.Contains(logs.String(), "owned-guest list unavailable") {
+		t.Fatalf("a failed ownership read must act on nothing: calls %q", r2.calls)
+	}
+}
+
+func TestReportJSONShape(t *testing.T) {
+	now := at(7, 10, 30)
+	tr, _, _ := newT(t, &fakeRunner{out: measuredOut}, running(9201), &backup.InFlight{}, &now)
+	tr.Pass(context.Background())
+	b, err := json.Marshal(tr.GuestDiskTrimStatus(context.Background()))
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, k := range []string{`"schedule":`, `"guests":[{"vmid":9201`, `"last_attempt_at":"2026-10-07T08:30:00Z"`, `"ok":true`,
+		`"bytes_trimmed":90143313920`, `"mounts":2`, `"duration_seconds":`, `"last_ok_at":"2026-10-07T08:30:00Z"`} {
+		if !strings.Contains(string(b), k) {
+			t.Errorf("report JSON lacks %s: %s", k, b)
+		}
+	}
+	if strings.Contains(string(b), `"error"`) {
+		t.Errorf("an ok result must omit error: %s", b)
+	}
+}

internal/guesthook/install.go

@@ -38,35 +38,26 @@ const snippetBody = `#!/bin/sh
 exit 0
 `
 
-// InstallSnippet writes the pre-start hook wrapper into the PVE snippets dir (idempotent, root-owned,
-// executable). The agent runs as a non-root service user, so it writes an agent-writable temp file then
-// `install`s it host-root (same pattern as the bootstrap mount + dnsmasq drop-ins). Safe to call repeatedly.
-// The temp file is a RANDOM-named os.CreateTemp (audit B1): a fixed, predictable /tmp name could be
-// pre-created by another local user and rewritten between our write and root's install (TOCTOU into a
-// root-executed hookscript). The final mode comes from `install -m`, so the 0600 temp is fine.
-func InstallSnippet(ctx context.Context, runner proxmox.Runner) error {
-	f, err := os.CreateTemp("", "felhom-guest-hook-*.sh")
+// SnippetReady (R-861, agent v0.146.0) reports whether the pre-start hook is in place: a regular file at path whose
+// content is exactly snippetBody. The hook is a FIXED, ROOT-OWNED file that arrives with the signed config bundle
+// (configs/felhom-guest-hook.sh, pinned byte-identical by TestSnippetEqualsTheBundle). The agent no longer installs it:
+// until v0.146.0 it `install`ed it from /tmp, and Proxmox runs a hookscript as root at every guest start — so the
+// install grant was a root shell for a compromised agent. A missing or different hook is an error the caller logs; it
+// then does NOT register the hook (a guest whose hookscript is missing does not start).
+func SnippetReady(path string) error {
+	fi, err := os.Lstat(path)
 	if err != nil {
-		return fmt.Errorf("guesthook: create temp snippet: %w", err)
+		return fmt.Errorf("guesthook: %s is missing — it arrives with the signed config bundle (agent_config_update): %w", path, err)
 	}
-	tmp := f.Name()
-	defer os.Remove(tmp)
-	if _, err := f.WriteString(snippetBody); err != nil {
-		f.Close()
-		return fmt.Errorf("guesthook: write temp snippet: %w", err)
+	if !fi.Mode().IsRegular() {
+		return fmt.Errorf("guesthook: %s is not a regular file", path)
 	}
-	if err := f.Close(); err != nil {
-		return fmt.Errorf("guesthook: close temp snippet: %w", err)
+	b, err := os.ReadFile(path)
+	if err != nil {
+		return fmt.Errorf("guesthook: read %s: %w", path, err)
 	}
-	// Ensure the snippets dir exists FIRST (B2, DRILL-day0-cleanroom-2026-07-03): a fresh PVE has
-	// no /var/lib/vz/snippets, and `install` (without -D) won't create the parent — the whole
-	// hook install silently failed on a freshly-bootstrapped box. Fenced root op like the install
-	// itself; idempotent.
-	if _, stderr, err := runner.Run(ctx, "mkdir", "-p", SnippetDir); err != nil {
-		return fmt.Errorf("guesthook: ensure snippets dir %s: %w: %s", SnippetDir, err, string(stderr))
-	}
-	if _, stderr, err := runner.Run(ctx, "install", "-m", "0755", "--", tmp, SnippetPath); err != nil {
-		return fmt.Errorf("guesthook: install snippet to %s: %w: %s", SnippetPath, err, string(stderr))
+	if string(b) != snippetBody {
+		return fmt.Errorf("guesthook: %s differs from this agent's hook — the next config bundle replaces it", path)
 	}
 	return nil
 }

internal/guesthook/install_test.go

@@ -4,7 +4,7 @@ import (
 	"context"
 	"io"
 	"os"
-	"regexp"
+	"path/filepath"
 	"testing"
 )
 
@@ -29,100 +29,34 @@ func (r *recordingRunner) RunStdin(ctx context.Context, _ io.Reader, name string
 	return r.Run(ctx, name, args...)
 }
 
-// TestInstallSnippet_RandomTempName is the audit-B1 negative test: the staged install SOURCE must be a
-// RANDOM os.CreateTemp name (felhom-guest-hook-<random>.sh), never the fixed, pre-creatable
-// /tmp/felhom-guest-hook.sh (a local TOCTOU into a root-executed hookscript), and two consecutive
-// installs must stage through DIFFERENT paths.
-func TestInstallSnippet_RandomTempName(t *testing.T) {
-	r := &recordingRunner{}
-	if err := InstallSnippet(context.Background(), r); err != nil {
-		t.Fatalf("InstallSnippet #1: %v", err)
+// R-861 (agent v0.146.0): the hook file comes with the signed config bundle; the agent never installs it, only checks.
+// RED-PROOF (audits/hub-safety-2026-10-05/partF/red-proof.txt): make SnippetReady accept any content → the
+// "differs" case fails.
+func TestSnippetReady(t *testing.T) {
+	d := t.TempDir()
+	p := filepath.Join(d, "felhom-guest-hook.sh")
+	if err := SnippetReady(p); err == nil {
+		t.Fatal("a missing hook read as ready — the guest would get a hookscript that does not exist")
 	}
-	if err := InstallSnippet(context.Background(), r); err != nil {
-		t.Fatalf("InstallSnippet #2: %v", err)
+	_ = os.WriteFile(p, []byte("#!/bin/sh\nid > /tmp/x\n"), 0o755)
+	if err := SnippetReady(p); err == nil {
+		t.Fatal("a hook with other content read as ready")
 	}
-	var installs [][]string
-	for _, call := range r.calls {
-		if call[0] == "install" {
-			installs = append(installs, call)
-		}
+	_ = os.WriteFile(p, []byte(snippetBody), 0o755)
+	if err := SnippetReady(p); err != nil {
+		t.Fatalf("the bundle's hook was not accepted: %v", err)
 	}
-	if len(installs) != 2 {
-		t.Fatalf("expected 2 install calls, got %d: %v", len(installs), r.calls)
-	}
-
-	randomName := regexp.MustCompile(`felhom-guest-hook-[^/\\]+\.sh$`)
-	fixedName := regexp.MustCompile(`felhom-guest-hook\.sh$`)
-	var srcs []string
-	for i, call := range installs {
-		// install -m 0755 -- <src> <dest>
-		if len(call) != 6 {
-			t.Fatalf("call %d: unexpected vector %v", i, call)
-		}
-		src, dest := call[4], call[5]
-		if dest != SnippetPath {
-			t.Errorf("call %d: dest = %q, want %q", i, dest, SnippetPath)
-		}
-		if !randomName.MatchString(src) {
-			t.Errorf("call %d: source %q does not match the random felhom-guest-hook-*.sh pattern", i, src)
-		}
-		if fixedName.MatchString(src) {
-			t.Errorf("call %d: source %q is the FIXED predictable temp name (B1 TOCTOU)", i, src)
-		}
-		srcs = append(srcs, src)
-	}
-	if srcs[0] == srcs[1] {
-		t.Errorf("two consecutive installs staged through the SAME source path %q — must be random per call", srcs[0])
-	}
-	// Non-hollow: the staged file must actually carry the snippet body at install time.
-	for i, c := range r.srcContent {
-		if c != snippetBody {
-			t.Errorf("call %d: staged content is not the snippet body (got %d bytes)", i, len(c))
-		}
-	}
-	// And the temp is cleaned up after.
-	for _, src := range srcs {
-		if _, err := os.Stat(src); err == nil {
-			t.Errorf("staged temp %q left behind (defer os.Remove missing)", src)
-		}
+	l := filepath.Join(d, "link.sh")
+	_ = os.Symlink(p, l)
+	if err := SnippetReady(l); err == nil {
+		t.Fatal("a symlink read as the hook")
 	}
 }
 
-// B2 Scenario D (DRILL-day0-cleanroom-2026-07-03): on a fresh PVE, /var/lib/vz/snippets does not
-// exist and `install` (no -D) cannot create it — the drill saw
-// `install: cannot create regular file … No such file or directory` and the guest silently got no
-// pre-start self-heal hook. InstallSnippet must therefore issue a `mkdir -p <SnippetDir>` fenced op
-// BEFORE the `install` op. Pre-fix wrong outcome: no mkdir call at all — only the doomed install.
-func TestInstallSnippet_EnsuresSnippetsDirFirst(t *testing.T) {
-	r := &recordingRunner{}
-	if err := InstallSnippet(context.Background(), r); err != nil {
-		t.Fatalf("InstallSnippet: %v", err)
-	}
-
-	mkdirIdx, installIdx := -1, -1
-	for i, call := range r.calls {
-		switch call[0] {
-		case "mkdir":
-			if mkdirIdx == -1 {
-				mkdirIdx = i
-				want := []string{"mkdir", "-p", SnippetDir}
-				if len(call) != 3 || call[1] != want[1] || call[2] != want[2] {
-					t.Errorf("mkdir vector = %v, want %v (the sudoers fence matches exactly this argv)", call, want)
-				}
-			}
-		case "install":
-			if installIdx == -1 {
-				installIdx = i
-			}
-		}
-	}
-	if mkdirIdx == -1 {
-		t.Fatalf("no `mkdir -p %s` op issued — on a fresh box the snippet install fails ENOENT (B2); calls: %v", SnippetDir, r.calls)
-	}
-	if installIdx == -1 {
-		t.Fatalf("no install op issued; calls: %v", r.calls)
-	}
-	if mkdirIdx > installIdx {
-		t.Fatalf("mkdir (call %d) must PRECEDE install (call %d) — order: %v", mkdirIdx, installIdx, r.calls)
+// The bundle's copy is byte-identical to the body the agent checks against.
+func TestSnippetEqualsTheBundle(t *testing.T) {
+	got, err := os.ReadFile(filepath.Join("..", "..", "configs", "felhom-guest-hook.sh"))
+	if err != nil || string(got) != snippetBody {
+		t.Fatalf("configs/felhom-guest-hook.sh differs from snippetBody (err %v)", err)
 	}
 }

internal/hub/collect.go

@@ -10,6 +10,7 @@ import (
 	"log/slog"
 	"os"
 	"strings"
+	"sync"
 	"time"
 
 	"gitea.dooplex.hu/admin/felhom-agent/internal/capability"
@@ -89,6 +90,12 @@ type GuestNetReporter interface {
 	GuestNetStatus(ctx context.Context) *GuestNetStatus
 }
 
+// GuestDiskTrimReporter is the R-444 seam the weekly trim job plugs into (same consumer-side pattern — hub does not
+// import fstrim). nil (feature not wired) → no guest_disk_trim stanza.
+type GuestDiskTrimReporter interface {
+	GuestDiskTrimStatus(ctx context.Context) *GuestDiskTrimStatus
+}
+
 // Collector builds a HostReport from read-only sources. All deps are behind narrow
 // interfaces for unit testing.
 type Collector struct {
@@ -107,6 +114,7 @@ type Collector struct {
 	pbsdr            PBSDRReporter                                 // slice 2: PBS DR tier bridge state (nil → stanza omitted)
 	ctrlSup          ControllerSupervisorReporter                  // R-523: in-guest controller supervisor (nil → stanza omitted)
 	guestNet         GuestNetReporter                              // R-54: per-guest network watchdog (nil → stanza omitted)
+	diskTrim         GuestDiskTrimReporter                         // R-444: weekly guest disk trim (nil → stanza omitted)
 	selfUpdate       SelfUpdateReporter                            // D1: agent self-update pending status (nil → false)
 	mgmtPlane        MgmtPlaneReporter                             // G1: management-plane health (nil → stanza omitted)
 	oob              OOBReporter                                   // H1: operator-access health (nil → stanza omitted)
@@ -115,6 +123,7 @@ type Collector struct {
 	backupTarget     func() ConfiguredBackupTarget                 // R-109: primary backup tier id (nil → recipe records unknown)
 	hostID           string
 	agentVersion     string
+	selfSHA          func() string // R-349: sha256 of the running binary; default runningBinarySHA256
 	logger           *slog.Logger
 	now              func() time.Time
 }
@@ -135,6 +144,7 @@ func NewCollector(px proxmoxReader, cf CloudflaredProber, storage StorageObserve
 		temp:         SysfsTempReader{}, // slice 9: real sysfs reader by default; tests inject a fake
 		hostID:       hostID,
 		agentVersion: agentVersion,
+		selfSHA:      runningBinarySHA256,
 		logger:       logger,
 		now:          func() time.Time { return time.Now().UTC() },
 	}
@@ -218,6 +228,12 @@ func (c *Collector) SetGuestNetReporter(g GuestNetReporter) *Collector {
 	return c
 }
 
+// SetGuestDiskTrimReporter wires the R-444 weekly trim job as a report source (nil-safe → stanza omitted).
+func (c *Collector) SetGuestDiskTrimReporter(r GuestDiskTrimReporter) *Collector {
+	c.diskTrim = r
+	return c
+}
+
 // SelfUpdateReporter is the D1 seam the selfupdate commit-manager plugs into (same consumer-side
 // pattern — hub does not import selfupdate). nil (feature not wired) → pending=false on the report.
 type SelfUpdateReporter interface {
@@ -337,6 +353,7 @@ func (c *Collector) Collect(ctx context.Context) (*HostReport, error) {
 		HostID:       c.hostID,
 		ReportedAt:   c.now().Format(time.RFC3339),
 		AgentVersion: c.agentVersion,
+		AgentSHA256:  c.agentSHA256(),
 		Host:         host,
 		Guests:       c.collectGuests(ctx),
 		// storage_targets populated this slice (slice 5) via the observer; the rest stay
@@ -373,6 +390,10 @@ func (c *Collector) Collect(ctx context.Context) (*HostReport, error) {
 	if c.guestNet != nil {
 		report.GuestNet = c.guestNet.GuestNetStatus(ctx)
 	}
+	// R-444: the last weekly trim result per guest (nil reporter = not wired → stanza omitted).
+	if c.diskTrim != nil {
+		report.GuestDiskTrim = c.diskTrim.GuestDiskTrimStatus(ctx)
+	}
 	// D1: agent self-update pending status (nil reporter → pending=false, the steady state).
 	if c.selfUpdate != nil {
 		report.SelfUpdatePending, report.SelfUpdatePendingVersion = c.selfUpdate.SelfUpdatePending()
@@ -431,8 +452,28 @@ const pbsWrapperPath = "/usr/local/sbin/felhom-pbs-apply"
 // unreadable file yields "", which the hub reads as UNKNOWN rather than as drift — a host that
 // legitimately has no DR wrapper must not light up amber. The file is 0755, so no privilege is
 // needed to read it.
-func pbsWrapperSHA256() string {
-	f, err := os.Open(pbsWrapperPath)
+func pbsWrapperSHA256() string { return fileSHA256(pbsWrapperPath) }
+
+// selfExePath is the running binary as the kernel holds it. /proc/self/exe, not the installed path:
+// after an A/B flip the file at /usr/local/bin/felhom-agent may already be the NEXT binary while this
+// process still runs the old one, and the report must describe what runs (R-349). Test seam.
+var selfExePath = "/proc/self/exe"
+
+// runningBinarySHA256 hashes the running binary ONCE per process — the bytes cannot change under a
+// running process, and re-hashing ~20 MB every report cycle buys nothing. A failed read is cached as
+// "" (UNKNOWN); it never fails the report.
+var runningBinarySHA256 = sync.OnceValue(func() string { return fileSHA256(selfExePath) })
+
+func (c *Collector) agentSHA256() string {
+	if c.selfSHA == nil {
+		return ""
+	}
+	return c.selfSHA()
+}
+
+// fileSHA256 is the hex sha256 of a file's bytes, or "" when it cannot be read.
+func fileSHA256(path string) string {
+	f, err := os.Open(path)
 	if err != nil {
 		return ""
 	}

internal/hub/collect_agentsha_test.go

@@ -0,0 +1,64 @@
+package hub
+
+import (
+	"context"
+	"crypto/sha256"
+	"encoding/hex"
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+// R-349: the report carries the sha256 of the binary that is RUNNING, so the hub can tell a
+// hand-built proof binary from the vouched artifact of the same version string. The consequence
+// asserted: the wire field equals the hash of this very test binary's bytes (read independently via
+// os.Executable, a different channel from /proc/self/exe), and it is on the wire as agent_sha256.
+func TestCollect_AgentSHA256IsTheRunningBinary(t *testing.T) {
+	exe, err := os.Executable()
+	if err != nil {
+		t.Skipf("os.Executable: %v", err)
+	}
+	raw, err := os.ReadFile(exe)
+	if err != nil {
+		t.Fatalf("read own binary: %v", err)
+	}
+	sum := sha256.Sum256(raw)
+	want := hex.EncodeToString(sum[:])
+
+	px := &fakePx{node: "n", ns: newTestNodeStatus()}
+	c := NewCollector(px, fakeProber{status: "running"}, nil, nil, nil, nil, "h", "0.3.0", quietLogger())
+	r, err := c.Collect(context.Background())
+	if err != nil {
+		t.Fatalf("Collect: %v", err)
+	}
+	if r.AgentSHA256 != want {
+		t.Fatalf("agent_sha256 = %q, want the running binary's %q", r.AgentSHA256, want)
+	}
+	b, err := json.Marshal(r)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !strings.Contains(string(b), `"agent_sha256":"`+want+`"`) {
+		t.Fatalf("agent_sha256 not on the wire: %s", b)
+	}
+}
+
+// An unreadable binary is UNKNOWN (empty, omitted) — never a made-up hash, never a failed report.
+func TestFileSHA256_UnreadableIsEmpty(t *testing.T) {
+	if got := fileSHA256(filepath.Join(t.TempDir(), "absent")); got != "" {
+		t.Fatalf("absent file hashed to %q, want empty", got)
+	}
+	px := &fakePx{node: "n", ns: newTestNodeStatus()}
+	c := NewCollector(px, fakeProber{status: "running"}, nil, nil, nil, nil, "h", "0.3.0", quietLogger())
+	c.selfSHA = func() string { return "" }
+	r, err := c.Collect(context.Background())
+	if err != nil {
+		t.Fatalf("Collect must not fail on an unreadable binary: %v", err)
+	}
+	b, _ := json.Marshal(r)
+	if strings.Contains(string(b), "agent_sha256") {
+		t.Fatalf("empty agent_sha256 must be omitted: %s", b)
+	}
+}

internal/hub/collect_disktrim_test.go

@@ -0,0 +1,54 @@
+package hub
+
+import (
+	"context"
+	"encoding/json"
+	"testing"
+)
+
+// R-444: the guest_disk_trim stanza must reach a report built through the PRODUCTION collect path, be absent from
+// the wire when the job is not wired, and carry the keys the hub's System page reads.
+
+type fakeDiskTrim struct{ st *GuestDiskTrimStatus }
+
+func (f fakeDiskTrim) GuestDiskTrimStatus(context.Context) *GuestDiskTrimStatus { return f.st }
+
+func TestCollect_GuestDiskTrim(t *testing.T) {
+	px := &fakePx{node: "n", ns: newTestNodeStatus()}
+	c := NewCollector(px, fakeProber{status: "running", detail: "connected"}, fakeObserver{}, nil, nil, nil, "h", "0.150.0", quietLogger())
+	r, err := c.Collect(context.Background())
+	if err != nil {
+		t.Fatalf("Collect: %v", err)
+	}
+	b, _ := json.Marshal(r)
+	var m map[string]any
+	_ = json.Unmarshal(b, &m)
+	if _, ok := m["guest_disk_trim"]; ok {
+		t.Fatalf("guest_disk_trim on the wire with no reporter wired: %s", b)
+	}
+
+	c.SetGuestDiskTrimReporter(fakeDiskTrim{st: &GuestDiskTrimStatus{Schedule: "weekly", Guests: []GuestDiskTrim{{
+		VMID: 9201, LastAttemptAt: "2026-10-07T08:30:00Z", OK: true, BytesTrimmed: 90143313920, Mounts: 2,
+		DurationSeconds: 24.4, LastOKAt: "2026-10-07T08:30:00Z",
+	}}}})
+	r, err = c.Collect(context.Background())
+	if err != nil {
+		t.Fatalf("Collect: %v", err)
+	}
+	b, _ = json.Marshal(r)
+	m = nil
+	_ = json.Unmarshal(b, &m)
+	dt, ok := m["guest_disk_trim"].(map[string]any)
+	if !ok || dt["schedule"] != "weekly" {
+		t.Fatalf("guest_disk_trim missing or wrong on the wire: %s", b)
+	}
+	g := dt["guests"].([]any)[0].(map[string]any)
+	for _, k := range []string{"vmid", "last_attempt_at", "ok", "bytes_trimmed", "mounts", "duration_seconds", "last_ok_at"} {
+		if _, ok := g[k]; !ok {
+			t.Fatalf("guest_disk_trim.guests[0] lacks %q: %v", k, g)
+		}
+	}
+	if g["bytes_trimmed"] != float64(90143313920) || g["ok"] != true {
+		t.Fatalf("values did not survive the round trip: %v", g)
+	}
+}

internal/hub/dr_recipe.go

@@ -54,11 +54,13 @@ const (
 	DRReasonNoPBSStorage = "no_pbs_storage_observed"
 )
 
-// PBSRootNamespace is how the recipe spells PBS's root namespace. The PBS API spells it as the EMPTY
-// string (and `pct restore --ns root` would name a namespace that does not exist) — "root" is a display
-// convention this wire has always used, kept here so the field's meaning did not change under R-106.
-// Only a box with no `namespace` line in its pbs storage.cfg stanza ever emits it.
-const PBSRootNamespace = "root"
+// PBSRootNamespace is how the recipe spells PBS's root namespace: the EMPTY string, PBS's own spelling (R-124,
+// agent v0.147.0). It used to be the display word "root", which no PBS namespace is named — an operator pasting it
+// into `proxmox-backup-client … --ns root` during a real recovery got a failure. An empty namespace is ambiguous on
+// its own, so READ IT WITH namespace_state: resolved + "" = the root namespace (pass no --ns, or --ns ""); unknown +
+// "" = the agent could not tell. Only a box with no `namespace` line in its pbs storage.cfg stanza emits it.
+// Pinned by TestDRRecipe_PBSNamespaceRootIsResolvedNotUnknown and TestR124_RootNamespaceOnTheWireIsPBSSpelling.
+const PBSRootNamespace = ""
 
 // DRRecipeHostHalf is the agent-emitted half (guest/drive/storage/PBS scaffolding). Derived entirely
 // from facts the report already collects — no new privileged reads.
@@ -123,8 +125,8 @@ type DRPBSCoord struct {
 	RepoID string `json:"repo_id"` // the PVE pbs storage id (e.g. "felhom-pbs") — not a token
 	// Namespace is the PBS namespace the restore targets, resolved from the pbs storage's storage.cfg
 	// stanza — the same field `vzdump --storage <pbs>` makes PVE read, so the recipe cannot disagree
-	// with the backup that produced the snapshot. PBSRootNamespace when the box has no namespace
-	// configured; "" when NamespaceState is unknown.
+	// with the backup that produced the snapshot. PBSRootNamespace ("", PBS's spelling, R-124) when the box has no
+	// namespace configured; also "" when NamespaceState is unknown — consult NamespaceState.
 	//
 	// R-106: this used to come from the listed snapshot's own `ns`, which PBS does not echo per item once
 	// the request is already namespace-scoped via `?ns=` (internal/pbs/client.go). The field was

internal/hub/dr_recipe_test.go

@@ -63,8 +63,8 @@ func TestBuildDRRecipeHostHalf(t *testing.T) {
 		t.Error("felhom-flash (local-dir user-data drive) missing from drives")
 	}
 	// pbs: latest snapshot's coords + the pbs storage id as repo_id.
-	if h.PBS == nil || h.PBS.RepoID != "felhom-pbs" || h.PBS.Namespace != "root" || h.PBS.LatestSnapshotID != "9201" {
-		t.Errorf("pbs coord = %+v, want repo felhom-pbs/root/9201", h.PBS)
+	if h.PBS == nil || h.PBS.RepoID != "felhom-pbs" || h.PBS.Namespace != PBSRootNamespace || h.PBS.LatestSnapshotID != "9201" {
+		t.Errorf("pbs coord = %+v, want repo felhom-pbs, the root namespace (\"\", R-124), snapshot 9201", h.PBS)
 	}
 }
 
@@ -252,7 +252,7 @@ func TestDRRecipe_PBSNamespaceIsThePerCustomerOne(t *testing.T) {
 }
 
 // TestDRRecipe_PBSNamespaceRootIsResolvedNotUnknown: a box with a pbs storage and NO namespace line is
-// genuinely in the root namespace. That is an answer, not a gap — it must read resolved/"root", so the
+// genuinely in the root namespace. That is an answer, not a gap — it must read resolved/"" (PBS's spelling, R-124), so the
 // honest root case is never confused with "I could not tell".
 func TestDRRecipe_PBSNamespaceRootIsResolvedNotUnknown(t *testing.T) {
 	h := BuildDRRecipeHostHalf(nil,
@@ -453,3 +453,39 @@ func assertNoSecretKeys(t *testing.T, jsonBytes []byte) {
 	}
 	walk("<root>", v)
 }
+
+// R-124: on the WIRE the root namespace is PBS's own spelling — an empty string, present (not omitted), beside
+// namespace_state "resolved". The display word "root" names no PBS namespace, and `--ns root` fails in a recovery.
+// RED-PROOF: set PBSRootNamespace back to "root" → this test fails.
+func TestR124_RootNamespaceOnTheWireIsPBSSpelling(t *testing.T) {
+	h := BuildDRRecipeHostHalf(nil,
+		[]StorageTarget{{Name: "felhom-pbs", Type: StorageTypePBS, Content: "backup", PBSNamespace: ""}},
+		capturedDemoFelhomSnapshots(),
+		ConfiguredBackupTarget{StorageID: "felhom-pbs", Known: true})
+	b, err := json.Marshal(h.PBS)
+	if err != nil {
+		t.Fatal(err)
+	}
+	var m map[string]any
+	if err := json.Unmarshal(b, &m); err != nil {
+		t.Fatal(err)
+	}
+	ns, present := m["namespace"]
+	if !present {
+		t.Fatalf("namespace key missing from %s — an omitted key reads as 'unknown', not 'root'", b)
+	}
+	if ns != "" {
+		t.Fatalf("root namespace on the wire = %q, want \"\" (PBS's spelling; no namespace is named %q)", ns, ns)
+	}
+	if m["namespace_state"] != DRStateResolved {
+		t.Fatalf("namespace_state = %v, want %q beside the empty root namespace", m["namespace_state"], DRStateResolved)
+	}
+	// A configured namespace still passes through unchanged.
+	h2 := BuildDRRecipeHostHalf(nil,
+		[]StorageTarget{{Name: "felhom-pbs", Type: StorageTypePBS, Content: "backup", PBSNamespace: "demo-felhom"}},
+		capturedDemoFelhomSnapshots(),
+		ConfiguredBackupTarget{StorageID: "felhom-pbs", Known: true})
+	if h2.PBS.Namespace != "demo-felhom" {
+		t.Fatalf("configured namespace = %q, want demo-felhom", h2.PBS.Namespace)
+	}
+}

internal/hub/report.go

@@ -18,6 +18,16 @@ type HostReport struct {
 	HostID       string `json:"host_id"`     // echoes config.Hub.HostID
 	ReportedAt   string `json:"reported_at"` // RFC3339, agent clock
 	AgentVersion string `json:"agent_version"`
+	// AgentSHA256 is the sha256 of the binary this process is RUNNING (read through /proc/self/exe,
+	// once per process), R-349. The version string cannot tell a hand-built proof binary from the
+	// published, vouched artifact of the same version — same source, different bytes (`-trimpath
+	// -buildvcs=false` in release-agent.sh) — so self-update sees "already installed" and never
+	// corrects it. Reporting the bytes lets the hub compare against the vouched agent_sha256, the
+	// same mechanism host.wrapper_sha256 is for the PBS wrapper (R-50b(a)).
+	//
+	// Empty = unreadable, which the hub must treat as UNKNOWN, never as drift. Pinned by
+	// TestCollect_AgentSHA256IsTheRunningBinary.
+	AgentSHA256 string `json:"agent_sha256,omitempty"`
 
 	Host   HostMetrics `json:"host"`
 	Guests []Guest     `json:"guests"`
@@ -114,6 +124,11 @@ type HostReport struct {
 	// on HostReport would have been the only report block named against that convention.
 	GuestNet *GuestNetStatus `json:"guest_net,omitempty"`
 
+	// GuestDiskTrim is the weekly guest disk trim stanza (R-444, `09` §3 decision 139): the schedule and, per owned
+	// guest, the LAST trim result as persisted by the agent (it survives a restart). Present only when the trim job is
+	// wired; an empty `guests` list means the job runs and no guest has been trimmed yet. No secret.
+	GuestDiskTrim *GuestDiskTrimStatus `json:"guest_disk_trim,omitempty"`
+
 	// LogTail is the agent's on-demand debug-ring tail (v0.83.0 observability) — the agent
 	// mirror of the controller's report log_tails channel. Present ONLY on the heartbeat
 	// right after the control envelope requested it (log_tail_requested); consume-once on
@@ -198,6 +213,27 @@ type GuestNetGuest struct {
 	Message       string `json:"message,omitempty"`
 }
 
+// GuestDiskTrimStatus is the R-444 weekly trim stanza. `schedule` is a plain description of when the job runs (local
+// time of the host); `guests` holds one entry per owned guest that has had at least one trim attempt.
+type GuestDiskTrimStatus struct {
+	Schedule string          `json:"schedule"`
+	Guests   []GuestDiskTrim `json:"guests,omitempty"`
+}
+
+// GuestDiskTrim is one guest's LAST trim attempt. `ok` with `last_attempt_at` is the verdict of that attempt — never
+// read the time alone as success; `last_ok_at` is the last attempt that succeeded ("" = never). `bytes_trimmed` is
+// the sum of the "(N bytes) trimmed" lines `pct fstrim` printed, over `mounts` mount points.
+type GuestDiskTrim struct {
+	VMID            int     `json:"vmid"`
+	LastAttemptAt   string  `json:"last_attempt_at"`
+	OK              bool    `json:"ok"`
+	BytesTrimmed    int64   `json:"bytes_trimmed"`
+	Mounts          int     `json:"mounts"`
+	DurationSeconds float64 `json:"duration_seconds"`
+	LastOKAt        string  `json:"last_ok_at,omitempty"`
+	Error           string  `json:"error,omitempty"`
+}
+
 type PBSDRStatus struct {
 	State          string `json:"state"`
 	StorageID      string `json:"storage_id,omitempty"`
@@ -418,6 +454,17 @@ type SmartSummary struct {
 	ReallocatedSectors   *int `json:"reallocated_sectors"`
 	PendingSectors       *int `json:"pending_sectors"`
 	OfflineUncorrectable *int `json:"offline_uncorrectable"`
+	// R-330 (disk health Phase 2): three more SATA raw counters. omitempty + pointer: absent (an
+	// older agent, an NVMe/USB device, or a drive that does not report the attribute) is OMITTED —
+	// unknown, never a zero (S-39). Wire only: no verdict reads them yet.
+	//   187 Reported_Uncorrect — the failing drive's most telling counter (normalized 1 vs thresh 0,
+	//       raw 1001) while SMART still said PASSED.
+	//   188 Command_Timeout — some vendors PACK several counters into the 48-bit raw value, so the
+	//       number is carried as reported and must not be compared across vendors.
+	//   199 UDMA_CRC_Error_Count — cabling / link errors, not the medium.
+	ReportedUncorrect *int64 `json:"reported_uncorrect,omitempty"`
+	CommandTimeout    *int64 `json:"command_timeout,omitempty"`
+	UDMACRCErrors     *int64 `json:"udma_crc_errors,omitempty"`
 
 	// NVMe attributes.
 	CriticalWarning *int `json:"critical_warning"`

internal/lanresolver/ensure_dnsmasq_test.go

@@ -0,0 +1,122 @@
+package lanresolver
+
+import (
+	"context"
+	"io"
+	"log/slog"
+	"os"
+	"path/filepath"
+	"strings"
+	"sync"
+	"testing"
+)
+
+// recRunner records every privileged command EnsureDnsmasq would run and succeeds — nothing reaches
+// apt, systemctl or the root checker.
+type recRunner struct {
+	mu    sync.Mutex
+	calls []string
+}
+
+func (r *recRunner) Run(_ context.Context, name string, args ...string) ([]byte, []byte, error) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	r.calls = append(r.calls, strings.Join(append([]string{name}, args...), " "))
+	return nil, nil, nil
+}
+
+func (r *recRunner) RunStdin(ctx context.Context, _ io.Reader, name string, args ...string) ([]byte, []byte, error) {
+	return r.Run(ctx, name, args...)
+}
+
+func (r *recRunner) installed() bool {
+	for _, c := range r.calls {
+		if strings.HasPrefix(c, "apt-get install") && strings.HasSuffix(c, " dnsmasq") {
+			return true
+		}
+	}
+	return false
+}
+
+// fixtureRoot builds a fake host root holding exactly the given relative files and points the REAL
+// probe at it for the test's duration.
+func fixtureRoot(t *testing.T, files ...string) {
+	t.Helper()
+	root := t.TempDir()
+	for _, f := range files {
+		p := filepath.Join(root, f)
+		if err := os.MkdirAll(filepath.Dir(p), 0o755); err != nil {
+			t.Fatal(err)
+		}
+		if err := os.WriteFile(p, nil, 0o644); err != nil {
+			t.Fatal(err)
+		}
+	}
+	prev := hostRoot
+	hostRoot = root
+	t.Cleanup(func() { hostRoot = prev })
+}
+
+func ensure(t *testing.T) *recRunner {
+	t.Helper()
+	r := &recRunner{}
+	m := NewManager(r, "192.0.2.10", []string{"1.1.1.1"}, slog.New(slog.NewTextHandler(io.Discard, nil)))
+	if err := m.EnsureDnsmasq(context.Background()); err != nil {
+		t.Fatalf("EnsureDnsmasq: %v", err)
+	}
+	return r
+}
+
+// R-317: a host with `dnsmasq-base` (the /usr/sbin/dnsmasq binary) but WITHOUT the `dnsmasq` package
+// (the service unit) must get the package installed — else the following `systemctl enable --now
+// dnsmasq` hits a unit that does not exist and LAN name resolution silently never comes up.
+//
+// RED-PROOF: probe "usr/sbin/dnsmasq" instead of the unit paths in dnsmasqUnitInstalled → this fails
+// with "install was skipped".
+func TestEnsureDnsmasq_BinaryWithoutUnitInstalls(t *testing.T) {
+	fixtureRoot(t, "usr/sbin/dnsmasq")
+	r := ensure(t)
+	if !r.installed() {
+		t.Fatalf("install was skipped on a dnsmasq-base-only host (binary present, unit absent) — "+
+			"the enable that follows targets a missing unit (R-317). calls: %q", r.calls)
+	}
+}
+
+func TestEnsureDnsmasq_UnitPresentSkipsInstall(t *testing.T) {
+	for _, unit := range []string{"usr/lib/systemd/system/dnsmasq.service", "lib/systemd/system/dnsmasq.service"} {
+		t.Run(unit, func(t *testing.T) {
+			fixtureRoot(t, "usr/sbin/dnsmasq", unit)
+			if r := ensure(t); r.installed() {
+				t.Fatalf("apt-get install ran although the dnsmasq unit is present at %s: %q", unit, r.calls)
+			}
+		})
+	}
+}
+
+func TestEnsureDnsmasq_NothingPresentInstalls(t *testing.T) {
+	fixtureRoot(t)
+	if r := ensure(t); !r.installed() {
+		t.Fatalf("install skipped on a host with no dnsmasq at all: %q", r.calls)
+	}
+}
+
+// Production wiring for the hostRoot seam: the shipped probe resolves against the real root and asks
+// about the unit the `dnsmasq` package owns — never the dnsmasq-base binary.
+func TestEnsureDnsmasq_ProductionProbeIsTheUnit(t *testing.T) {
+	if hostRoot != "/" {
+		t.Fatalf("hostRoot default = %q, want \"/\" — the production probe would look in the wrong tree", hostRoot)
+	}
+	var sawUsrLib bool
+	for _, p := range dnsmasqUnitPaths {
+		full := filepath.Join(hostRoot, p)
+		if strings.HasSuffix(full, "/sbin/dnsmasq") || strings.HasSuffix(full, "/bin/dnsmasq") {
+			t.Errorf("probe path %s is the dnsmasq-base binary, not the dnsmasq unit (R-317)", full)
+		}
+		if full == "/usr/lib/systemd/system/dnsmasq.service" {
+			sawUsrLib = true
+		}
+	}
+	if !sawUsrLib {
+		t.Errorf("probe paths %q miss /usr/lib/systemd/system/dnsmasq.service (dpkg -S: owned by dnsmasq)", dnsmasqUnitPaths)
+	}
+}

internal/lanresolver/lanresolver.go

@@ -33,6 +33,8 @@ const (
 	DropinDir = "/etc/dnsmasq.d"
 	// BaseDropin holds the host-wide listen/upstream config (one per host).
 	BaseDropin = "felhom-resolver-base.conf"
+	// PrivApply is the root content checker that installs a drop-in (R-861).
+	PrivApply = "/usr/local/sbin/felhom-priv-apply"
 )
 
 // RenderBase returns the host-wide dnsmasq drop-in: bind to the host LAN IP (+ loopback), no-resolv,
@@ -99,11 +101,35 @@ func NewManager(runner proxmox.Runner, hostIP string, upstreams []string, logger
 	}
 }
 
+// hostRoot is the filesystem root the install probe resolves against: "/" in production; a test
+// points it at a fixture tree so the REAL probe runs against files it controls.
+var hostRoot = "/"
+
+// dnsmasqUnitPaths are where the `dnsmasq` package ships its systemd unit (Debian; /lib is the
+// pre-usrmerge spelling). R-317: probe the UNIT, never /usr/sbin/dnsmasq — that binary belongs to
+// `dnsmasq-base`, so a host carrying dnsmasq-base without dnsmasq used to skip the install and then
+// `systemctl enable --now dnsmasq` failed against a unit that is not there (resolver never up).
+// Pinned by TestEnsureDnsmasq_BinaryWithoutUnitInstalls.
+var dnsmasqUnitPaths = []string{
+	"usr/lib/systemd/system/dnsmasq.service",
+	"lib/systemd/system/dnsmasq.service",
+}
+
+// dnsmasqUnitInstalled reports whether the dnsmasq service unit (the `dnsmasq` package) is present.
+func dnsmasqUnitInstalled() bool {
+	for _, p := range dnsmasqUnitPaths {
+		if _, err := os.Stat(filepath.Join(hostRoot, p)); err == nil {
+			return true
+		}
+	}
+	return false
+}
+
 // EnsureDnsmasq makes dnsmasq present + enabled and writes the host base config. Idempotent: it
 // installs the package only when absent, and writes the base drop-in only when its content changes.
 func (m *Manager) EnsureDnsmasq(ctx context.Context) error {
-	if _, err := os.Stat("/usr/sbin/dnsmasq"); err != nil { // metadata read, no privilege needed
-		m.logger.Info("lanresolver: dnsmasq absent — installing")
+	if !dnsmasqUnitInstalled() { // metadata read, no privilege needed
+		m.logger.Info("lanresolver: dnsmasq service unit absent — installing")
 		if out, errOut, ierr := m.runner.Run(ctx, "apt-get", "install", "-y", "-q", "dnsmasq"); ierr != nil {
 			return fmt.Errorf("install dnsmasq: %s: %w", strings.TrimSpace(string(errOut))+string(out), ierr)
 		}
@@ -263,7 +289,13 @@ func (m *Manager) writeFileIfChanged(ctx context.Context, path, content, mode st
 		return false, fmt.Errorf("write temp: %w", err)
 	}
 	tmp.Close()
-	if _, errOut, err := m.runner.Run(ctx, "install", "-m", mode, tmpName, path); err != nil {
+	// R-861 (v0.146.0): a dnsmasq drop-in reaches /etc/dnsmasq.d only through the root checker, which allows exactly
+	// the lines RenderBase/RenderGuestDropin write (a `dhcp-script=` would run as root). Mode is fixed (0644) there.
+	_ = mode
+	if filepath.Dir(path) != DropinDir {
+		return false, fmt.Errorf("drop-in %s is not in %s", path, DropinDir)
+	}
+	if _, errOut, err := m.runner.Run(ctx, PrivApply, "dnsmasq", tmpName, filepath.Base(path)); err != nil {
 		return false, fmt.Errorf("install %s: %s: %w", path, strings.TrimSpace(string(errOut)), err)
 	}
 	return true, nil

internal/lanresolver/r861_privapply_contract_test.go

@@ -0,0 +1,22 @@
+package lanresolver
+
+import (
+	"strings"
+	"testing"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/privapplytest"
+)
+
+// R-861: both drop-ins the resolver renders are accepted by the root checker; a dhcp-script line is not.
+func TestPrivApply_AcceptsTheRenderedDropins(t *testing.T) {
+	if got := privapplytest.Check(t, "dnsmasq", BaseDropin, RenderBase("192.168.0.104", []string{"1.1.1.1", "8.8.8.8"})); got != "OK" {
+		t.Errorf("base drop-in: %s", got)
+	}
+	if got := privapplytest.Check(t, "dnsmasq", DropinName("demo-hp"), RenderGuestDropin("demo-hp", "enkisfelhom.hu", "192.168.0.138")); got != "OK" {
+		t.Errorf("guest drop-in: %s", got)
+	}
+	evil := RenderGuestDropin("demo-hp", "enkisfelhom.hu", "192.168.0.138") + "dhcp-script=/var/lib/felhom-agent/x\n"
+	if got := privapplytest.Check(t, "dnsmasq", "felhom-x.conf", evil); !strings.HasPrefix(got, "REFUSED") {
+		t.Fatalf("control: dhcp-script was not refused: %s", got)
+	}
+}

internal/localapi/crashguard.go

@@ -0,0 +1,102 @@
+package localapi
+
+import (
+	"encoding/json"
+	"errors"
+	"io"
+	"io/fs"
+	"net/http"
+	"os"
+	"time"
+)
+
+// GET /host/crash-guard (R-856, `09` §3 decision 143): what the host's crash guard
+// (configs/felhom-crash-guard, `11` §5.9) recorded about the most recent HOST boot. The controller
+// reads it once after it starts: when the host's last boot followed an UNCLEAN stop, its app mails
+// wait ~15 minutes instead of the normal 90 s boot grace.
+//
+// Read-only and Proxmox-free: the agent reads the guard's state file (root-owned, 0644 — the
+// non-root agent can read it) and passes four fields through. Host-wide, token-authed (any valid
+// per-guest token sees the host's view, as GET /host/metrics does).
+//
+// NEVER an error page. A missing file (no guard installed, or no boot recorded yet), an unreadable
+// one, or one that does not parse answers 200 with present:false — the controller reads that as
+// UNKNOWN and keeps its normal boot grace. Pinned by TestR856_CrashGuard*.
+
+// defaultCrashGuardStatePath is where configs/felhom-crash-guard writes its state (STATE_DIR there).
+const defaultCrashGuardStatePath = "/var/lib/felhom-crash-guard/state.json"
+
+// crashGuardStateMax bounds the read; the real file is well under 4 KiB.
+const crashGuardStateMax = 1 << 20
+
+// CrashGuardResponse is the data block of GET /host/crash-guard. Field names are the controller's
+// agentapi.CrashGuardState (felhom-controller internal/agentapi/crashguard.go) — a wire contract,
+// pinned by TestR856_CrashGuardWireMatchesControllerClient.
+type CrashGuardResponse struct {
+	Present         bool   `json:"present"`
+	LastBootAt      string `json:"last_boot_at,omitempty"` // RFC3339 UTC ("2006-01-02T15:04:05Z")
+	LastBootUnclean bool   `json:"last_boot_unclean"`
+	Tripped         bool   `json:"tripped"`
+}
+
+// crashGuardFile is the subset of the guard's state.json the route passes through. Every other key
+// (armed, boot_id, config, unclean_boots, last_trip, ...) is ignored.
+type crashGuardFile struct {
+	LastBootAt      string `json:"last_boot_at"`
+	LastBootUnclean bool   `json:"last_boot_unclean"`
+	Tripped         bool   `json:"tripped"`
+}
+
+// readCrashGuardState reads and parses the guard's state file. ok=false on ANY failure (missing,
+// unreadable, oversized, not a JSON object, a field of the wrong type); reason says which, for the log.
+func readCrashGuardState(path string) (resp CrashGuardResponse, ok bool, reason string) {
+	f, err := os.Open(path)
+	if err != nil {
+		if errors.Is(err, fs.ErrNotExist) {
+			return resp, false, "no state file"
+		}
+		return resp, false, "unreadable: " + err.Error()
+	}
+	defer f.Close()
+	raw, err := io.ReadAll(io.LimitReader(f, crashGuardStateMax+1))
+	if err != nil {
+		return resp, false, "read: " + err.Error()
+	}
+	if len(raw) > crashGuardStateMax {
+		return resp, false, "state file too large"
+	}
+	var st crashGuardFile
+	// Unmarshal into a struct fails on a non-object top level (null decodes, so reject it below).
+	if err := json.Unmarshal(raw, &st); err != nil {
+		return resp, false, "unparseable: " + err.Error()
+	}
+	var probe map[string]json.RawMessage
+	if err := json.Unmarshal(raw, &probe); err != nil || probe == nil {
+		return resp, false, "unparseable: not a JSON object"
+	}
+	resp = CrashGuardResponse{Present: true, LastBootUnclean: st.LastBootUnclean, Tripped: st.Tripped}
+	// Normalise to RFC3339 UTC; an unparseable time passes through as-is (the controller reads an
+	// unparseable boot time as "not this start's boot" → its normal grace).
+	if t, perr := time.Parse(time.RFC3339, st.LastBootAt); perr == nil {
+		resp.LastBootAt = t.UTC().Format(time.RFC3339)
+	} else {
+		resp.LastBootAt = st.LastBootAt
+	}
+	return resp, true, ""
+}
+
+func (s *Server) handleCrashGuard(w http.ResponseWriter, r *http.Request, vmid int) {
+	path := s.crashGuardStatePath
+	if path == "" {
+		path = defaultCrashGuardStatePath
+	}
+	resp, ok, reason := readCrashGuardState(path)
+	if !ok {
+		s.logger.Debug("local-api: /host/crash-guard not present", "vmid", vmid, "reason", reason)
+		writeOK(w, CrashGuardResponse{Present: false})
+		return
+	}
+	s.logger.Debug("local-api: /host/crash-guard served", "vmid", vmid,
+		"last_boot_at", resp.LastBootAt, "last_boot_unclean", resp.LastBootUnclean, "tripped", resp.Tripped)
+	writeOK(w, resp)
+}

internal/localapi/crashguard_test.go

@@ -0,0 +1,205 @@
+package localapi
+
+import (
+	"encoding/json"
+	"io"
+	"log/slog"
+	"net/http"
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+// The shape of /var/lib/felhom-crash-guard/state.json as read on demo-hp on 2026-10-06 (values from
+// that read where they matter; lists/objects kept to the same key set).
+const crashGuardFixture = `{
+  "armed": true,
+  "boot_id": "3f1c0f1e-6a0b-4d7e-9b7a-0c2d4e6f8a1b",
+  "config": {"LIMIT": 3, "WINDOW_MINUTES": 60, "PANIC_SECONDS": 10},
+  "kernel_panic": 10,
+  "last_boot_at": "2026-10-05T07:56:41Z",
+  "last_boot_unclean": true,
+  "last_trip": {},
+  "rearmed_at": "2026-10-04T14:02:11Z",
+  "rearmed_by": "operator",
+  "tripped": false,
+  "unclean_boots": ["2026-10-05T07:56:41Z"],
+  "unclean_boots_24h": 1,
+  "unclean_boots_in_window": 1,
+  "updated_at": "2026-10-05T07:57:02Z",
+  "version": 1
+}`
+
+// controllerCrashGuardState is a COPY of the controller's wire type, felhom-controller
+// controller/internal/agentapi/crashguard.go `CrashGuardState` (commit 8b13a5e) — same field names,
+// same tags. If either side renames a key, the contract test below fails.
+type controllerCrashGuardState struct {
+	Present         bool   `json:"present"`
+	LastBootAt      string `json:"last_boot_at,omitempty"`
+	LastBootUnclean bool   `json:"last_boot_unclean"`
+	Tripped         bool   `json:"tripped"`
+}
+
+func newCrashGuardServer(t *testing.T, statePath string) http.Handler {
+	t.Helper()
+	srv, err := NewServer(Options{
+		ListenAddr: "127.0.0.1:0",
+		Guests:     &fakeGuests{},
+		Backups:    &fakeBackups{},
+		Store:      &fakeStore{},
+		Storage:    fakeStorage{},
+		Tokens:     staticTokens{"A": 8200, "B": 9300},
+		Logger:     slog.New(slog.NewTextHandler(io.Discard, nil)),
+	})
+	if err != nil {
+		t.Fatalf("new server: %v", err)
+	}
+	srv.crashGuardStatePath = statePath
+	return srv.Handler()
+}
+
+func writeCrashGuardFixture(t *testing.T, body string) string {
+	t.Helper()
+	p := filepath.Join(t.TempDir(), "state.json")
+	if err := os.WriteFile(p, []byte(body), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	return p
+}
+
+// getCrashGuard calls the route and decodes the envelope with the CONTROLLER's type.
+func getCrashGuard(t *testing.T, h http.Handler, token string) (int, controllerCrashGuardState, string) {
+	t.Helper()
+	w := do(t, h, "GET", "/host/crash-guard", token, "")
+	var env struct {
+		OK   bool                      `json:"ok"`
+		Data controllerCrashGuardState `json:"data"`
+	}
+	if w.Code == http.StatusOK {
+		if err := json.Unmarshal(w.Body.Bytes(), &env); err != nil {
+			t.Fatalf("decode %q: %v", w.Body.String(), err)
+		}
+		if !env.OK {
+			t.Fatalf("ok=false: %s", w.Body.String())
+		}
+	}
+	return w.Code, env.Data, w.Body.String()
+}
+
+// A present state file (demo-hp's shape) passes the three facts through.
+func TestR856_CrashGuardPresentFile(t *testing.T) {
+	h := newCrashGuardServer(t, writeCrashGuardFixture(t, crashGuardFixture))
+	code, st, body := getCrashGuard(t, h, "A")
+	if code != http.StatusOK {
+		t.Fatalf("got %d, want 200 (%s)", code, body)
+	}
+	want := controllerCrashGuardState{Present: true, LastBootAt: "2026-10-05T07:56:41Z", LastBootUnclean: true, Tripped: false}
+	if st != want {
+		t.Fatalf("state = %+v, want %+v", st, want)
+	}
+
+	// A tripped, clean boot reads back as such (both bools are carried, not defaulted).
+	h = newCrashGuardServer(t, writeCrashGuardFixture(t,
+		`{"last_boot_at":"2026-10-05T09:56:41+02:00","last_boot_unclean":false,"tripped":true,"version":1}`))
+	_, st, _ = getCrashGuard(t, h, "B")
+	want = controllerCrashGuardState{Present: true, LastBootAt: "2026-10-05T07:56:41Z", LastBootUnclean: false, Tripped: true}
+	if st != want {
+		t.Fatalf("offset time / tripped: state = %+v, want %+v (time normalised to UTC Z)", st, want)
+	}
+}
+
+// No state file (no guard on this host, or no boot recorded yet) → 200 present:false.
+func TestR856_CrashGuardMissingFile(t *testing.T) {
+	h := newCrashGuardServer(t, filepath.Join(t.TempDir(), "absent", "state.json"))
+	code, st, body := getCrashGuard(t, h, "A")
+	if code != http.StatusOK {
+		t.Fatalf("missing file: got %d, want 200 (%s)", code, body)
+	}
+	if st.Present || st.LastBootUnclean || st.Tripped || st.LastBootAt != "" {
+		t.Fatalf("missing file: state = %+v, want present:false and nothing else", st)
+	}
+}
+
+// A garbled file → 200 present:false, never a 5xx — every shape of garbage.
+func TestR856_CrashGuardGarbageFile(t *testing.T) {
+	for name, body := range map[string]string{
+		"truncated":  crashGuardFixture[:40],
+		"not json":   "this is not json\n",
+		"empty":      "",
+		"null":       "null",
+		"array":      `[{"last_boot_unclean":true}]`,
+		"wrong type": `{"last_boot_at":"2026-10-05T07:56:41Z","last_boot_unclean":"yes","tripped":false}`,
+		"lone brace": "{",
+	} {
+		t.Run(name, func(t *testing.T) {
+			h := newCrashGuardServer(t, writeCrashGuardFixture(t, body))
+			code, st, raw := getCrashGuard(t, h, "A")
+			if code != http.StatusOK {
+				t.Fatalf("got %d, want 200 (%s)", code, raw)
+			}
+			if st.Present || st.LastBootUnclean {
+				t.Fatalf("garbage %q read as %+v, want present:false", name, st)
+			}
+		})
+	}
+	// The path is a directory, not a file: unreadable → present:false, 200.
+	h := newCrashGuardServer(t, t.TempDir())
+	if code, st, raw := getCrashGuard(t, h, "A"); code != http.StatusOK || st.Present {
+		t.Fatalf("directory path: got %d %+v (%s), want 200 present:false", code, st, raw)
+	}
+}
+
+// No / unknown token → 401, like every sibling route; a cross-guest ?vmid= → 403.
+func TestR856_CrashGuardRequiresGuestToken(t *testing.T) {
+	h := newCrashGuardServer(t, writeCrashGuardFixture(t, crashGuardFixture))
+	for _, tok := range []string{"", "bogus"} {
+		w := do(t, h, "GET", "/host/crash-guard", tok, "")
+		if w.Code != http.StatusUnauthorized {
+			t.Fatalf("token %q: got %d, want 401", tok, w.Code)
+		}
+		if json.Valid(w.Body.Bytes()) {
+			var env struct {
+				Data controllerCrashGuardState `json:"data"`
+			}
+			_ = json.Unmarshal(w.Body.Bytes(), &env)
+			if env.Data.Present || env.Data.LastBootUnclean {
+				t.Fatalf("token %q: the refusal leaked the state: %s", tok, w.Body.String())
+			}
+		}
+	}
+	if w := do(t, h, "GET", "/host/crash-guard?vmid=9300", "A", ""); w.Code != http.StatusForbidden {
+		t.Fatalf("cross-guest query: got %d, want 403", w.Code)
+	}
+}
+
+// Wire contract: every key the controller's CrashGuardState decodes is emitted under exactly that
+// name, and the agent emits no key the controller does not know.
+func TestR856_CrashGuardWireMatchesControllerClient(t *testing.T) {
+	h := newCrashGuardServer(t, writeCrashGuardFixture(t, crashGuardFixture))
+	w := do(t, h, "GET", "/host/crash-guard", "A", "")
+	var env struct {
+		OK   bool                       `json:"ok"`
+		Data map[string]json.RawMessage `json:"data"`
+	}
+	if err := json.Unmarshal(w.Body.Bytes(), &env); err != nil || !env.OK {
+		t.Fatalf("envelope: %v %s", err, w.Body.String())
+	}
+	want := []string{"present", "last_boot_at", "last_boot_unclean", "tripped"}
+	for _, k := range want {
+		if _, ok := env.Data[k]; !ok {
+			t.Errorf("agent does not emit %q, which the controller decodes (%s)", k, w.Body.String())
+		}
+	}
+	if len(env.Data) != len(want) {
+		t.Errorf("agent emits %d keys, controller knows %d: %s", len(env.Data), len(want), w.Body.String())
+	}
+	// And the agent's own type agrees with the controller's copy, field for field.
+	var mine CrashGuardResponse
+	var theirs controllerCrashGuardState
+	raw, _ := json.Marshal(env.Data)
+	_ = json.Unmarshal(raw, &mine)
+	_ = json.Unmarshal(raw, &theirs)
+	if (controllerCrashGuardState{mine.Present, mine.LastBootAt, mine.LastBootUnclean, mine.Tripped}) != theirs {
+		t.Errorf("agent %+v vs controller %+v", mine, theirs)
+	}
+}

internal/localapi/disks.go

@@ -398,9 +398,16 @@ func (s *Server) handleDisks(w http.ResponseWriter, r *http.Request, vmid int) {
 						di.Smart = &sm
 					}
 				}
-				if total, used, okc := statfsCapacity(d.MountPath); okc {
-					di.TotalBytes, di.UsedBytes = total, used
-					di.UsedFraction = float64(used) / float64(total)
+				// R-118: statfs ONLY while the drive's device is present. With the device gone the raw
+				// mountpoint reverts to a bare directory on the ROOT filesystem, and statfs would report
+				// pve-root's size as this drive's (measured: a 4 GB drive advertising 46 GiB). Same trap
+				// observe.go guards on the Observe path. Absent → capacity left zero (unknown), never root's.
+				// Pinned by TestDisks_UnionPath_AbsentDeviceReportsNoRootCapacity.
+				if s.devicePresent(d.MountPath) {
+					if total, used, okc := statfsCapacity(d.MountPath); okc {
+						di.TotalBytes, di.UsedBytes = total, used
+						di.UsedFraction = float64(used) / float64(total)
+					}
 				}
 				out = append(out, di)
 			}
@@ -761,6 +768,11 @@ type FormatResponse struct {
 	// signature — the customer authorizes the wipe of their own data drive.
 	NeedsConfirmation bool   `json:"needs_confirmation,omitempty"`
 	DurableID         string `json:"durable_id,omitempty"` // the durable id to confirm against (user-data)
+	// FSUUID (on Formatted) is the UUID of the filesystem the agent just made, verified against the
+	// bound durable id after mkfs (R-25). The caller mounts THIS — re-resolving a UUID from the /dev
+	// path later can name another disk if /dev re-enumerated. "" = not verified: the caller must not
+	// substitute a path-resolved guess silently.
+	FSUUID string `json:"fs_uuid,omitempty"`
 	// PendingOp is set on a SYSTEM/BACKUP data-bearing refusal — the exact op the operator must sign.
 	PendingOp *PendingOp `json:"pending_op,omitempty"`
 }
@@ -808,7 +820,7 @@ func (s *Server) handleDiskFormatStatus(w http.ResponseWriter, r *http.Request,
 	writeOK(w, map[string]any{
 		"vmid": vmid, "phase": job.Phase, "device": job.Device, "fstype": job.FSType,
 		"durable_id": job.DurableID, "error": job.Error, "started_at": job.StartedAt, "updated_at": job.UpdatedAt,
-		"job_id": job.JobID,
+		"job_id": job.JobID, "fs_uuid": job.FSUUID, // R-25: "" until done + verified
 	})
 }
 
@@ -871,7 +883,7 @@ func (s *Server) handleDiskFormat(w http.ResponseWriter, r *http.Request, vmid i
 				"format refused (device may have changed since inspection): "+rerr.Error())
 			return
 		}
-		done := s.startFormatDetached(device, blankDurable, req.FSType, true)
+		job, done := s.startFormatDetached(device, blankDurable, req.FSType, true)
 		if err := s.awaitFormat(r.Context(), done, vmid, device); err != nil {
 			if err == errFormatClientGone {
 				return // client gone; mkfs continues detached + the job record records the outcome
@@ -880,7 +892,7 @@ func (s *Server) handleDiskFormat(w http.ResponseWriter, r *http.Request, vmid i
 			writeErr(w, http.StatusBadGateway, "format failed: "+err.Error())
 			return
 		}
-		writeOK(w, FormatResponse{VMID: vmid, Device: device, Formatted: true, DataBearing: false, DurableID: blankDurable, Reason: "blank device formatted " + req.FSType})
+		writeOK(w, FormatResponse{VMID: vmid, Device: device, Formatted: true, DataBearing: false, DurableID: blankDurable, FSUUID: job.FSUUID, Reason: "blank device formatted " + req.FSType})
 		return
 	}
 
@@ -917,7 +929,7 @@ func (s *Server) handleDiskFormat(w http.ResponseWriter, r *http.Request, vmid i
 		// F20-BUG3: run the destructive mkfs DETACHED off s.baseCtx (bound durable id recorded for
 		// restart-recovery), so a request/client deadline can never SIGKILL it mid-write and corrupt the
 		// disk. We still wait to return the synchronous result (backward-compatible with the controller).
-		done := s.startFormatDetached(device, deviceDurable, req.FSType, false)
+		job, done := s.startFormatDetached(device, deviceDurable, req.FSType, false)
 		if err := s.awaitFormat(r.Context(), done, vmid, device); err != nil {
 			if err == errFormatClientGone {
 				return // client gone; the wipe continues detached + survives a restart via the job record
@@ -929,7 +941,7 @@ func (s *Server) handleDiskFormat(w http.ResponseWriter, r *http.Request, vmid i
 		s.logger.Warn("local-api: USER-DATA data-bearing format — CUSTOMER CONFIRMED (no operator signature)",
 			"vmid", vmid, "device", device, "durable_id", deviceDurable, "fstype", req.FSType, "why", probe.Reason())
 		writeOK(w, FormatResponse{VMID: vmid, Device: device, Formatted: true, DataBearing: true,
-			Role: string(role), DurableID: deviceDurable, Reason: "customer-confirmed wipe (" + probe.Reason() + ")"})
+			Role: string(role), DurableID: deviceDurable, FSUUID: job.FSUUID, Reason: "customer-confirmed wipe (" + probe.Reason() + ")"})
 		return
 	}
 

internal/localapi/disks_device_presence_test.go

@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"io"
 	"log/slog"
+	"runtime"
 	"strings"
 	"testing"
 
@@ -184,3 +185,39 @@ func TestDisks_DevicePresence_WireFieldIsFalseOnDeviceLoss(t *testing.T) {
 		t.Fatalf("the drive never reached the wire: %s", body)
 	}
 }
+
+// ── R-118 — an absent drive must not advertise the ROOT filesystem's capacity ───────────────────
+
+// TestDisks_UnionPath_AbsentDeviceReportsNoRootCapacity drives the REAL statfsCapacity (no capacity
+// seam): the registry drive's mount path is a real, bare temp directory — exactly what /mnt/<name>
+// becomes once its device is gone (a plain directory on the host's filesystem). With the device absent
+// the row must carry NO capacity; before R-118 the union path statfs'd that bare directory and reported
+// the host filesystem's size and usage as the drive's (46 GiB at 9.2 % for a 4 GB drive, measured).
+// The present half proves the test is not hollow: the same directory DOES yield capacity when the
+// device is there, so a zero on the absent half is the guard's doing, not a statfs failure.
+//
+// RED-PROOF: drop the `if s.devicePresent(d.MountPath)` guard around statfsCapacity in disks.go → the
+// absent subtest fails with "advertises ... bytes".
+func TestDisks_UnionPath_AbsentDeviceReportsNoRootCapacity(t *testing.T) {
+	if runtime.GOOS != "linux" {
+		t.Skip("statfsCapacity is linux-only; production target is linux")
+	}
+	bare := t.TempDir()
+	known := []storage.KnownTarget{
+		{Name: "cel", Type: hub.StorageTypeUSB, MountPath: bare, DurableID: "uuid:4242", UUID: "4242"},
+	}
+	t.Run("absent", func(t *testing.T) {
+		di := diskByMount(t, presenceServer(t, nil, known, true, false), bare)
+		if di.TotalBytes != 0 || di.UsedBytes != 0 || di.UsedFraction != 0 {
+			t.Errorf("absent drive advertises total=%d used=%d frac=%.3f — that is the filesystem UNDER "+
+				"the bare mountpoint, not the drive (R-118)", di.TotalBytes, di.UsedBytes, di.UsedFraction)
+		}
+	})
+	t.Run("present", func(t *testing.T) {
+		di := diskByMount(t, presenceServer(t, nil, known, true, true), bare)
+		if di.TotalBytes <= 0 {
+			t.Errorf("present drive reports no capacity (total=%d) — the guard over-corrected and the "+
+				"size bar is gone for every healthy registry drive", di.TotalBytes)
+		}
+	})
+}

internal/localapi/disks_test.go

@@ -28,6 +28,7 @@ type fakeDiskOps struct {
 	unmountCalls []string
 	candidates   []storage.CandidateDisk // returned by ListCandidateDisks
 	candErr      error
+	afterFormat  *storage.DeviceProbe // R-25: when set, InspectDevice returns it once a format ran
 }
 
 func (f *fakeDiskOps) ListCandidateDisks(_ context.Context) ([]storage.CandidateDisk, error) {
@@ -35,7 +36,12 @@ func (f *fakeDiskOps) ListCandidateDisks(_ context.Context) ([]storage.Candidate
 }
 
 func (f *fakeDiskOps) InspectDevice(_ context.Context, device string) (storage.DeviceProbe, error) {
+	f.mu.Lock()
 	p := f.probe
+	if f.afterFormat != nil && len(f.formatCalls) > 0 {
+		p = *f.afterFormat
+	}
+	f.mu.Unlock()
 	p.Device = device
 	return p, f.inspectErr
 }

internal/localapi/format_fsuuid_test.go

@@ -0,0 +1,96 @@
+package localapi
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"sync"
+	"testing"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/storage"
+)
+
+// R-25: the format answer carries the UUID of the filesystem the agent JUST made, verified against the
+// bound durable id after mkfs, so the controller mounts that filesystem rather than whatever the /dev
+// path resolves to a few requests later. The consequence asserted: the UUID on the wire (and in the
+// polled job record) is the new superblock's — and is EMPTY whenever the binding cannot be re-proved.
+
+const newFSUUID = "0fc63daf-8483-4772-8e79-3d69d8477de4"
+
+func confirmedFormat(t *testing.T, d *fakeDiskOps, srv *Server, fj *FormatJobStore) (string, *formatJob) {
+	t.Helper()
+	w := do(t, srv.Handler(), "POST", "/disks/format", "A", `{"device":"/dev/sdb1","fstype":"ext4","confirmed":true,"durable_id":"byid:wwn-/dev/sdb1"}`)
+	if w.Code != http.StatusOK {
+		t.Fatalf("confirmed format: %d (%s)", w.Code, w.Body.String())
+	}
+	var resp struct {
+		Data FormatResponse `json:"data"`
+	}
+	if err := json.Unmarshal(w.Body.Bytes(), &resp); err != nil {
+		t.Fatalf("decode: %v (%s)", err, w.Body.String())
+	}
+	if !resp.Data.Formatted {
+		t.Fatalf("not formatted: %s", w.Body.String())
+	}
+	return resp.Data.FSUUID, waitFormatPhase(t, fj, formatPhaseDone)
+}
+
+func confirmedGate() *fakeGate {
+	return &fakeGate{decision: WipeDecision{Allowed: true, Tier: "customer_confirmable", Reason: "customer_confirmed"}}
+}
+
+func TestFormat_ReportsNewFilesystemUUID(t *testing.T) {
+	d := &fakeDiskOps{probe: deviceProbeDataBearing(),
+		afterFormat: &storage.DeviceProbe{Probed: true, HasFilesystem: true, FSType: "ext4", FSUUID: newFSUUID}}
+	fj := tempFormatStore(t)
+	srv := formatServer(t, d, confirmedGate(), fj)
+
+	got, job := confirmedFormat(t, d, srv, fj)
+	if got != newFSUUID {
+		t.Fatalf("response fs_uuid = %q, want the new filesystem's %q", got, newFSUUID)
+	}
+	if job.FSUUID != newFSUUID {
+		t.Fatalf("job record fs_uuid = %q, want %q (the polled status path)", job.FSUUID, newFSUUID)
+	}
+}
+
+// The node moved between mkfs and the read-back: the bound durable id now resolves elsewhere. The
+// UUID must NOT be reported — reading it would name the other disk's filesystem.
+func TestFormat_FSUUIDWithheldWhenDurableIDMoved(t *testing.T) {
+	d := &fakeDiskOps{probe: deviceProbeDataBearing(),
+		afterFormat: &storage.DeviceProbe{Probed: true, HasFilesystem: true, FSType: "ext4", FSUUID: newFSUUID}}
+	fj := tempFormatStore(t)
+	srv := formatServer(t, d, confirmedGate(), fj)
+	var mu sync.Mutex
+	calls := 0
+	srv.reresolveWipe = func(_ context.Context, _ string) (string, error) {
+		mu.Lock()
+		defer mu.Unlock()
+		calls++
+		if calls == 1 {
+			return "/dev/sdb", nil // the pre-mkfs anti-retarget re-resolve
+		}
+		return "/dev/sdc", nil // after mkfs: the durable id now names another node
+	}
+
+	got, job := confirmedFormat(t, d, srv, fj)
+	if got != "" || job.FSUUID != "" {
+		t.Fatalf("fs_uuid reported after the durable id moved (response %q, job %q) — must be empty", got, job.FSUUID)
+	}
+	if calls < 2 {
+		t.Fatalf("the post-mkfs re-resolve never ran (calls=%d)", calls)
+	}
+}
+
+// The superblock did not read back as the requested filesystem → not verified → empty.
+func TestFormat_FSUUIDWithheldOnFSTypeMismatch(t *testing.T) {
+	d := &fakeDiskOps{probe: deviceProbeDataBearing(),
+		afterFormat: &storage.DeviceProbe{Probed: true, HasFilesystem: true, FSType: "xfs", FSUUID: newFSUUID}}
+	fj := tempFormatStore(t)
+	srv := formatServer(t, d, confirmedGate(), fj)
+
+	got, job := confirmedFormat(t, d, srv, fj)
+	if got != "" || job.FSUUID != "" {
+		t.Fatalf("fs_uuid reported for a superblock of the wrong type (response %q, job %q)", got, job.FSUUID)
+	}
+}

internal/localapi/formatjob.go

@@ -22,6 +22,10 @@ type formatJob struct {
 	Blank     bool   `json:"blank,omitempty"` // audit D3: blank (benign) format — recovery re-checks STILL-blank, not data-bearing
 	Phase     string `json:"phase"`           // running | done | failed
 	Error     string `json:"error,omitempty"`
+	// FSUUID is the filesystem UUID of the NEW filesystem, read by the agent right after mkfs on the
+	// device the bound durable id still resolves to (R-25). "" = not verified (the controller must not
+	// read that as a UUID). Set only on phase done.
+	FSUUID    string `json:"fs_uuid,omitempty"`
 	StartedAt string `json:"started_at"`
 	UpdatedAt string `json:"updated_at"`
 }
@@ -96,7 +100,10 @@ func (s *FormatJobStore) save(j *formatJob) error {
 // runs to completion and records the outcome. device is the ALREADY anti-retarget-resolved device; the
 // record carries durableID so a restart can re-resolve + re-run. blank marks a benign (blank-device)
 // format, so restart recovery re-checks STILL-blank rather than data-bearing (audit D3).
-func (s *Server) startFormatDetached(device, durableID, fstype string, blank bool) <-chan error {
+//
+// The returned job may be read (FSUUID) only AFTER a value arrives on done — the goroutine writes it
+// before the send, which is the happens-before edge.
+func (s *Server) startFormatDetached(device, durableID, fstype string, blank bool) (*formatJob, <-chan error) {
 	base := s.baseCtx
 	if base == nil {
 		base = context.Background()
@@ -116,10 +123,39 @@ func (s *Server) startFormatDetached(device, durableID, fstype string, blank boo
 		ctx, cancel := context.WithTimeout(base, 60*time.Minute)
 		defer cancel()
 		err := s.disks.Format(ctx, device, fstype)
+		if err == nil {
+			job.FSUUID = s.formattedFSUUID(ctx, device, durableID, fstype)
+		}
 		s.finishFormatJob(job, err)
 		done <- err
 	}()
-	return done
+	return job, done
+}
+
+// formattedFSUUID reads the UUID of the filesystem the agent has JUST made (R-25). The caller used to
+// re-resolve the UUID from the mutable /dev path afterwards, over separate requests — a re-enumeration
+// in that window could hand it ANOTHER disk's filesystem to mount. Here the bound durable id must still
+// resolve to the very device that was formatted (and re-derive to the same id), and the superblock must
+// carry the fstype that was asked for; anything else returns "" (not verified), never a guess.
+func (s *Server) formattedFSUUID(ctx context.Context, device, durableID, fstype string) string {
+	if durableID == "" || s.reresolveWipe == nil {
+		return ""
+	}
+	// The device now holds a filesystem, so the data-bearing anti-retarget re-resolve is the right one.
+	now, err := s.reresolveWipe(ctx, durableID)
+	if err != nil || now != device {
+		s.logger.Warn("format: new filesystem UUID NOT reported — bound durable id no longer resolves to the formatted device",
+			"device", device, "durable_id", durableID, "resolves_to", now, "err", err)
+		return ""
+	}
+	probe, err := s.disks.InspectDevice(ctx, device)
+	if err != nil || !probe.Probed || probe.FSType != fstype || probe.FSUUID == "" {
+		s.logger.Warn("format: new filesystem UUID NOT reported — superblock did not read back as the requested filesystem",
+			"device", device, "want_fstype", fstype, "got_fstype", probe.FSType, "has_uuid", probe.FSUUID != "", "err", err)
+		return ""
+	}
+	s.logger.Info("format: new filesystem bound to its durable id", "device", device, "durable_id", durableID, "fs_uuid", probe.FSUUID)
+	return probe.FSUUID
 }
 
 // finishFormatJob updates the persisted record to done/failed.
@@ -172,7 +208,7 @@ func (s *Server) RecoverFormatJob(ctx context.Context) {
 		return
 	}
 	s.logger.Warn("format-job recover: re-running interrupted format detached", "durable_id", job.DurableID, "device", device, "fstype", job.FSType, "blank", job.Blank)
-	_ = s.startFormatDetached(device, job.DurableID, job.FSType, job.Blank) // detached; updates the record on completion
+	_, _ = s.startFormatDetached(device, job.DurableID, job.FSType, job.Blank) // detached; updates the record on completion
 }
 
 // nowFn returns the server clock (testable), defaulting to time.Now.

internal/localapi/intermediary.go

@@ -121,32 +121,33 @@ func (b *GuestBinder) EnsureSharedParent(ctx context.Context) error {
 	// only the script (the unit was unchanged), so the earlier unit-only gate never redeployed it —
 	// leaving hosts running the pre-fix script (no make-private), whose self-bind stays in root's shared
 	// peer group and DOUBLES every drive bind. Comparing both files closes that deploy gap.
-	if sharedParentInstallStale(sharedParentUnitPath, sharedParentScriptPath) {
-		if ierr := b.installSharedParentUnit(ctx); ierr != nil {
-			b.logger.Warn("shared-parent: boot-persistence (re)install failed (live setup OK; survives until host reboot)", "err", ierr)
-		}
+	if err := b.ensureSharedParentBoot(ctx, sharedParentUnitPath, sharedParentScriptPath, sharedParentWantsLink); err != nil {
+		b.logger.Warn("shared-parent: boot persistence not enabled (live setup OK; survives until host reboot)", "err", err)
 	}
 	return nil
 }
 
-// stageTemp writes content to a fresh random-named temp file (os.CreateTemp pattern — `*` is replaced
-// by a random string) and returns its path. Caller removes it after the privileged `install`.
-func stageTemp(pattern, content string) (string, error) {
-	f, err := os.CreateTemp("", pattern)
-	if err != nil {
-		return "", err
+// sharedParentWantsLink exists once `systemctl enable felhom-shared-parent.service` ran (WantedBy=pve-guests.service).
+const sharedParentWantsLink = "/etc/systemd/system/pve-guests.service.wants/felhom-shared-parent.service"
+
+// ensureSharedParentBoot (R-861, agent v0.146.0) — the boot script and its unit are ROOT-OWNED FIXED files that arrive
+// with the signed config bundle (configs/felhom-shared-parent.{sh,service}, byte-identical to the constants above,
+// pinned by TestSharedParentFilesEqualTheBundle). The agent NEVER installs them any more: until v0.146.0 it installed
+// them from /tmp, and a script a root unit runs at boot was a root shell for a compromised agent. Here it only checks
+// them, and enables the unit (a fixed sudoers line) when the bundle has put it in place but nothing has enabled it — a
+// fresh install. Missing or different files: one warning, nothing run (the next bundle brings them).
+func (b *GuestBinder) ensureSharedParentBoot(ctx context.Context, unitPath, scriptPath, wantsLink string) error {
+	if sharedParentInstallStale(unitPath, scriptPath) {
+		return fmt.Errorf("%s or %s is missing or differs from this agent's — it arrives with the signed config bundle (agent_config_update); the agent does not install it (R-861)", scriptPath, unitPath)
 	}
-	name := f.Name()
-	if _, err := f.WriteString(content); err != nil {
-		f.Close()
-		os.Remove(name)
-		return "", err
+	if _, err := os.Lstat(wantsLink); err == nil {
+		return nil
 	}
-	if err := f.Close(); err != nil {
-		os.Remove(name)
-		return "", err
+	if err := b.run(ctx, "systemctl", "enable", "felhom-shared-parent.service"); err != nil {
+		return fmt.Errorf("enable unit: %w", err)
 	}
-	return name, nil
+	b.logger.Info("shared-parent: boot-persistence unit enabled (files from the config bundle)")
+	return nil
 }
 
 // sharedParentInstallStale reports whether the on-disk boot script OR unit is missing or differs from
@@ -162,37 +163,6 @@ func sharedParentInstallStale(unitPath, scriptPath string) bool {
 	return false
 }
 
-// installSharedParentUnit writes the script + unit (from agent-written temps) and enables the unit so the
-// shared parent is re-established on every host boot before pve-guests. Idempotent. The temps are
-// RANDOM-named os.CreateTemp files (audit B1): a fixed, predictable /tmp name could be pre-created by
-// another local user and rewritten between our write and root's install (TOCTOU into a root-executed
-// boot script). The final modes come from `install -m`, so the 0600 temps are fine.
-func (b *GuestBinder) installSharedParentUnit(ctx context.Context) error {
-	tmpScript, err := stageTemp("felhom-shared-parent-*.sh", sharedParentScript)
-	if err != nil {
-		return fmt.Errorf("write temp script: %w", err)
-	}
-	defer os.Remove(tmpScript)
-	if err := b.run(ctx, "install", "-m", "0755", "--", tmpScript, sharedParentScriptPath); err != nil {
-		return fmt.Errorf("install script: %w", err)
-	}
-	tmpUnit, err := stageTemp("felhom-shared-parent-*.service", sharedParentUnit)
-	if err != nil {
-		return fmt.Errorf("write temp unit: %w", err)
-	}
-	defer os.Remove(tmpUnit)
-	if err := b.run(ctx, "install", "-m", "0644", "--", tmpUnit, sharedParentUnitPath); err != nil {
-		return fmt.Errorf("install unit: %w", err)
-	}
-	if err := b.run(ctx, "systemctl", "daemon-reload"); err != nil {
-		return fmt.Errorf("daemon-reload: %w", err)
-	}
-	if err := b.run(ctx, "systemctl", "enable", "felhom-shared-parent.service"); err != nil {
-		return fmt.Errorf("enable unit: %w", err)
-	}
-	return nil
-}
-
 // AttachDrive binds a drive's felhom-data namespace under the stable parent so it appears live in the
 // guest at the returned stable path (via propagation — no pct, no reboot). `where` is the drive's RAW
 // host PVE mount (/mnt/<name>); only `<where>/felhom-data` crosses into the guest (confinement). The

internal/localapi/intermediary_install_test.go

@@ -4,94 +4,82 @@ import (
 	"context"
 	"io"
 	"os"
-	"regexp"
+	"path/filepath"
 	"testing"
 )
 
-// stagingRecorderRunner is a fake proxmox.Runner recording every call vector and snapshotting each
-// install SOURCE file's content at call time (the deferred os.Remove erases it afterwards).
-type stagingRecorderRunner struct {
-	calls      [][]string
-	srcContent map[string]string // install dest → staged source content
-}
+// R-861 (agent v0.146.0): the shared-parent boot script and unit arrive with the signed config bundle; the agent never
+// installs them. It checks them and enables the unit when nothing has.
+//
+// RED-PROOF (audits/hub-safety-2026-10-05/partF/red-proof.txt): put the old installSharedParentUnit call back (an
+// `install` of a /tmp file) → TestSharedParentBoot_NeverInstalls fails.
 
-func (r *stagingRecorderRunner) Run(_ context.Context, name string, args ...string) ([]byte, []byte, error) {
+type callRecorder struct{ calls [][]string }
+
+func (r *callRecorder) Run(_ context.Context, name string, args ...string) ([]byte, []byte, error) {
 	r.calls = append(r.calls, append([]string{name}, args...))
-	if name == "install" && len(args) >= 2 {
-		src, dest := args[len(args)-2], args[len(args)-1]
-		if r.srcContent == nil {
-			r.srcContent = map[string]string{}
-		}
-		b, _ := os.ReadFile(src)
-		r.srcContent[dest] = string(b)
-	}
 	return nil, nil, nil
 }
 
-func (r *stagingRecorderRunner) RunStdin(ctx context.Context, _ io.Reader, name string, args ...string) ([]byte, []byte, error) {
+func (r *callRecorder) RunStdin(ctx context.Context, _ io.Reader, name string, args ...string) ([]byte, []byte, error) {
 	return r.Run(ctx, name, args...)
 }
 
-// installSources returns the install-call source paths keyed by destination.
-func (r *stagingRecorderRunner) installSources() map[string][]string {
-	out := map[string][]string{}
-	for _, c := range r.calls {
-		if c[0] == "install" && len(c) >= 3 {
-			src, dest := c[len(c)-2], c[len(c)-1]
-			out[dest] = append(out[dest], src)
-		}
+func bootFiles(t *testing.T, script, unit string) (string, string, string) {
+	t.Helper()
+	d := t.TempDir()
+	sp, up := filepath.Join(d, "felhom-shared-parent.sh"), filepath.Join(d, "felhom-shared-parent.service")
+	if script != "" {
+		_ = os.WriteFile(sp, []byte(script), 0o755)
 	}
-	return out
+	if unit != "" {
+		_ = os.WriteFile(up, []byte(unit), 0o644)
+	}
+	return sp, up, filepath.Join(d, "wants-link")
 }
 
-// TestInstallSharedParent_RandomTempName is the audit-B1 negative test for the shared-parent boot
-// persistence install: both staged install SOURCES (script + unit) must be RANDOM os.CreateTemp names
-// (felhom-shared-parent-<random>.sh / .service), never the fixed, pre-creatable /tmp names (a local
-// TOCTOU into a root-executed boot script), and two consecutive installs must use DIFFERENT paths.
-func TestInstallSharedParent_RandomTempName(t *testing.T) {
-	r := &stagingRecorderRunner{}
+func TestSharedParentBoot_NeverInstalls(t *testing.T) {
+	for _, c := range []struct{ name, script, unit string }{
+		{"both missing", "", ""},
+		{"script differs", "#!/bin/sh\necho old\n", sharedParentUnit},
+		{"unit missing", sharedParentScript, ""},
+	} {
+		r := &callRecorder{}
+		b := NewGuestBinder(r, nil)
+		sp, up, link := bootFiles(t, c.script, c.unit)
+		if err := b.ensureSharedParentBoot(context.Background(), up, sp, link); err == nil {
+			t.Errorf("%s: no error — the operator would not learn the bundle is missing", c.name)
+		}
+		if len(r.calls) != 0 {
+			t.Errorf("%s: the agent ran %v — it must install nothing (R-861)", c.name, r.calls)
+		}
+	}
+}
+
+func TestSharedParentBoot_EnablesOnceTheBundleBroughtTheFiles(t *testing.T) {
+	r := &callRecorder{}
 	b := NewGuestBinder(r, nil)
-	if err := b.installSharedParentUnit(context.Background()); err != nil {
-		t.Fatalf("installSharedParentUnit #1: %v", err)
+	sp, up, link := bootFiles(t, sharedParentScript, sharedParentUnit)
+	if err := b.ensureSharedParentBoot(context.Background(), up, sp, link); err != nil {
+		t.Fatal(err)
 	}
-	if err := b.installSharedParentUnit(context.Background()); err != nil {
-		t.Fatalf("installSharedParentUnit #2: %v", err)
+	if len(r.calls) != 1 || len(r.calls[0]) != 3 || r.calls[0][0] != "systemctl" || r.calls[0][1] != "enable" ||
+		r.calls[0][2] != "felhom-shared-parent.service" {
+		t.Fatalf("want exactly `systemctl enable felhom-shared-parent.service`, got %v", r.calls)
 	}
+	_ = os.Symlink(up, link)
+	r.calls = nil
+	if err := b.ensureSharedParentBoot(context.Background(), up, sp, link); err != nil || len(r.calls) != 0 {
+		t.Fatalf("already enabled: want no calls, got %v (%v)", r.calls, err)
+	}
+}
 
-	srcs := r.installSources()
-	cases := []struct {
-		dest    string
-		random  *regexp.Regexp
-		fixed   *regexp.Regexp
-		content string
-	}{
-		{sharedParentScriptPath, regexp.MustCompile(`felhom-shared-parent-[^/\\]+\.sh$`),
-			regexp.MustCompile(`felhom-shared-parent\.sh$`), sharedParentScript},
-		{sharedParentUnitPath, regexp.MustCompile(`felhom-shared-parent-[^/\\]+\.service$`),
-			regexp.MustCompile(`felhom-shared-parent\.service$`), sharedParentUnit},
-	}
-	for _, tc := range cases {
-		got := srcs[tc.dest]
-		if len(got) != 2 {
-			t.Fatalf("dest %s: expected 2 install calls, got %d (%v)", tc.dest, len(got), got)
-		}
-		for i, src := range got {
-			if !tc.random.MatchString(src) {
-				t.Errorf("dest %s call %d: source %q does not match the random temp pattern", tc.dest, i, src)
-			}
-			if tc.fixed.MatchString(src) {
-				t.Errorf("dest %s call %d: source %q is the FIXED predictable temp name (B1 TOCTOU)", tc.dest, i, src)
-			}
-			if _, err := os.Stat(src); err == nil {
-				t.Errorf("dest %s: staged temp %q left behind (defer os.Remove missing)", tc.dest, src)
-			}
-		}
-		if got[0] == got[1] {
-			t.Errorf("dest %s: two consecutive installs staged through the SAME source %q — must be random per call", tc.dest, got[0])
-		}
-		// Non-hollow: the staged file carried the real content at install time.
-		if r.srcContent[tc.dest] != tc.content {
-			t.Errorf("dest %s: staged content mismatch (got %d bytes, want %d)", tc.dest, len(r.srcContent[tc.dest]), len(tc.content))
+// The bundle's copies are byte-identical to the constants the agent compares against.
+func TestSharedParentFilesEqualTheBundle(t *testing.T) {
+	for file, want := range map[string]string{"felhom-shared-parent.sh": sharedParentScript, "felhom-shared-parent.service": sharedParentUnit} {
+		got, err := os.ReadFile(filepath.Join("..", "..", "configs", file))
+		if err != nil || string(got) != want {
+			t.Errorf("configs/%s differs from the agent's constant (err %v)", file, err)
 		}
 	}
 }

internal/localapi/r894_saved_backup_test.go

@@ -0,0 +1,144 @@
+package localapi
+
+import (
+	"context"
+	"io"
+	"log/slog"
+	"net/http"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/backup"
+	"gitea.dooplex.hu/admin/felhom-agent/internal/hub"
+)
+
+// R-894 — after an agent restart, an UNREADABLE storage must fall back to the last success saved on
+// disk, not to "never". Measured 2026-10-05 on demo-hp: a restart at 04:57, the off-site storage
+// unreachable at 06:25, the 7-day tier (last copy 4 days old) read DUE, vzdump failed.
+//
+// Every test here builds a NEW server and a NEW BackupSuccessState from the same file — that is the
+// restart. The in-memory store (fakeStore) is always fresh, as after a real restart.
+
+// r894Server builds a two-tier server whose off-site tier answers the storage listing with lister.
+func r894Server(t *testing.T, path string, pbsSvc BackupService) *Server {
+	t.Helper()
+	srv, err := NewServer(Options{
+		ListenAddr: "127.0.0.1:0", Guests: &fakeGuests{}, Backups: &fakeBackups{}, Store: &fakeStore{},
+		Storage: fakeStorage{targets: []hub.StorageTarget{{Name: "local"}, {Name: "felhom-pbs"}}},
+		Tokens:  staticTokens{"A": 8200},
+		BackupTiers: []BackupTier{
+			{TargetID: "local", Cadence: 24 * time.Hour, Primary: true, Service: &fakeBackups{}},
+			{TargetID: "felhom-pbs", Cadence: 7 * 24 * time.Hour, Service: pbsSvc},
+		},
+		LastKnownBackups: backup.NewBackupSuccessState(path),
+		Logger:           slog.New(slog.NewTextHandler(io.Discard, nil)),
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	srv.baseCtx = context.Background()
+	srv.now = func() time.Time { return testNow }
+	return srv
+}
+
+// unreadable is the off-site storage as demo-hp saw it: "Can't connect to 10.77.0.1:8007".
+func unreadable() archiveLister {
+	return archiveLister{fakeBackups: &fakeBackups{}, err: errStorageRead}
+}
+
+// THE R-894 CASE, end to end. Agent 1 takes an off-site backup through POST /backup (the fake
+// runner's success is 12 h before testNow). The agent restarts. The storage cannot be read. The tier
+// must read NOT due, from the copy saved on disk.
+//
+// COMPANION RED-PROOF (observed): delete the `lookup == archiveUnknown && s.lastKnown != nil` block in
+// handleBackupDue → this fails with "after a restart an unreadable storage must fall back to the saved
+// copy (12 h old, 7-day tier) — NOT due; got {… Due:true … AgeState:unknown …}". Restored.
+func TestBackupDue_R894_RestartThenUnreadableStorage_FreshSavedCopyIsNotDue(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "backup-success-state.json")
+
+	// Agent 1: a real backup job through the endpoint the controller calls.
+	first := r894Server(t, path, &fakeBackups{})
+	if rr := do(t, first.Handler(), "POST", "/backup?target=felhom-pbs", "A", ""); rr.Code != http.StatusAccepted {
+		t.Fatalf("POST /backup: %d %s", rr.Code, rr.Body.String())
+	}
+	waitFor(t, func() bool {
+		_, ok := backup.NewBackupSuccessState(path).LastKnownSuccess("felhom-pbs", 8200)
+		return ok
+	})
+
+	// Agent 2: a restart (new server, new state from the same file), and the storage is unreachable.
+	got := dueFor(t, r894Server(t, path, unreadable()).Handler(), "felhom-pbs")
+	if got.Due {
+		t.Fatalf("after a restart an unreadable storage must fall back to the saved copy (12 h old, 7-day tier) — NOT due; got %+v", got)
+	}
+	if got.AgeState != AgeStateKnown || got.AgeSecs == nil || *got.AgeSecs != int64((12*time.Hour).Seconds()) {
+		t.Fatalf("the age must come from the saved copy (12 h, known); got %+v", got)
+	}
+}
+
+// The deliberate rule stays: an unreadable storage must not suppress a backup that IS due. A saved
+// copy older than the cadence reads DUE.
+//
+// COMPANION RED-PROOF (observed): make the fallback answer not-due whenever a saved copy exists
+// (`if fromDisk { …Due:false… }` before the cadence check) → this fails with "a saved copy 9 days old
+// under a 7-day cadence MUST read due". Restored.
+func TestBackupDue_R894_RestartThenUnreadableStorage_OldSavedCopyIsDue(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "backup-success-state.json")
+	st := backup.NewBackupSuccessState(path)
+	if err := st.RecordBackupSuccess("felhom-pbs", backupAt("felhom-pbs", 8200, 9*24*time.Hour, true)); err != nil {
+		t.Fatal(err)
+	}
+	got := dueFor(t, r894Server(t, path, unreadable()).Handler(), "felhom-pbs")
+	if !got.Due {
+		t.Fatalf("a saved copy 9 days old under a 7-day cadence MUST read due; got %+v", got)
+	}
+	if got.AgeState != AgeStateKnown {
+		t.Fatalf("the age is known (from disk); got %+v", got)
+	}
+}
+
+// No saved copy → the pre-R-894 answer, byte for byte: DUE, age UNKNOWN (never ABSENT — the controller
+// fires its window-gate valve only on absent, R-88).
+func TestBackupDue_R894_RestartThenUnreadableStorage_NoSavedCopyIsDueUnknown(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "backup-success-state.json")
+	got := dueFor(t, r894Server(t, path, unreadable()).Handler(), "felhom-pbs")
+	if !got.Due || got.AgeState != AgeStateUnknown || got.AgeSecs != nil {
+		t.Fatalf("no saved copy + unreadable storage must stay DUE with age unknown; got %+v", got)
+	}
+}
+
+// A storage that ANSWERS is the ground truth: an archive absent there makes the tier due even when the
+// file remembers a fresh success (a pruned or deleted copy must be made again).
+//
+// COMPANION RED-PROOF (observed): drop `lookup == archiveUnknown &&` from the fallback condition → this
+// fails with "the storage answered 'no archive' — the saved copy must NOT stand in for it". Restored.
+func TestBackupDue_R894_SavedCopyIgnoredWhenStorageAnswers(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "backup-success-state.json")
+	st := backup.NewBackupSuccessState(path)
+	if err := st.RecordBackupSuccess("felhom-pbs", backupAt("felhom-pbs", 8200, time.Hour, true)); err != nil {
+		t.Fatal(err)
+	}
+	absent := archiveLister{fakeBackups: &fakeBackups{}, found: false}
+	got := dueFor(t, r894Server(t, path, absent).Handler(), "felhom-pbs")
+	if !got.Due {
+		t.Fatalf("the storage answered 'no archive' — the saved copy must NOT stand in for it; got %+v", got)
+	}
+}
+
+// A FAILED backup is never saved: it must not make a tier look fresh after a restart.
+func TestBackupDue_R894_FailedBackupIsNotSaved(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "backup-success-state.json")
+	first := r894Server(t, path, &fakeBackups{failErr: "could not activate storage 'felhom-pbs'"})
+	if rr := do(t, first.Handler(), "POST", "/backup?target=felhom-pbs", "A", ""); rr.Code != http.StatusAccepted {
+		t.Fatalf("POST /backup: %d %s", rr.Code, rr.Body.String())
+	}
+	waitFor(t, func() bool { return len(first.store.Backups(context.Background())) == 1 })
+	if _, ok := backup.NewBackupSuccessState(path).LastKnownSuccess("felhom-pbs", 8200); ok {
+		t.Fatal("a failed backup must never be saved as a success")
+	}
+	got := dueFor(t, r894Server(t, path, unreadable()).Handler(), "felhom-pbs")
+	if !got.Due {
+		t.Fatalf("after a failed backup and a restart the tier must still be due; got %+v", got)
+	}
+}

internal/localapi/server.go

@@ -85,6 +85,13 @@ type BackupStore interface {
 	RestoreTests(ctx context.Context) []hub.RestoreTest
 }
 
+// LastKnownBackupStore (R-894) is the on-disk newest-success-per-tier record. Satisfied by
+// *backup.BackupSuccessState.
+type LastKnownBackupStore interface {
+	RecordBackupSuccess(target string, b hub.Backup) error
+	LastKnownSuccess(target string, vmid int) (time.Time, bool)
+}
+
 // StorageView yields the host's observed storage targets (for mapping a mount's storage id →
 // fast/slow class). Satisfied by *storage.Observer.
 type StorageView interface {
@@ -164,6 +171,10 @@ type Options struct {
 	// PRIMARY tier, inside the backup goroutine and BEFORE the host-wide heavy-op gate is released — so the OS leg
 	// that it starts can never overlap another backup or a restore-test (`11` C10). OPTIONAL — nil → nothing runs.
 	AfterPrimaryBackup func(ctx context.Context, vmid int)
+	// LastKnownBackups (R-894) keeps the newest successful backup per tier ON DISK, so the due-check's
+	// fallback for an UNREADABLE storage after an agent restart is the last known copy, not "never".
+	// nil = the pre-R-894 behaviour (in-memory record only).
+	LastKnownBackups LastKnownBackupStore
 	// Privileged runs the fenced root wrappers (E-2a: felhom-backup-target-apply). OPTIONAL — when
 	// nil, POST /backup/target reports "not configured". Satisfied by *proxmox.ExecRunner.
 	Privileged PrivilegedRunner
@@ -229,7 +240,6 @@ type Options struct {
 	// POST /escrow/recover-offsite-password. OPTIONAL — nil → that route reports "not configured"
 	// (503) instead of failing obscurely. Satisfied by escrow.OffsiteKeyRecoverer.
 	EscrowRecovery EscrowRecoverer
-
 }
 
 // defaultBackupCadence is the fallback /backup/due window when none is configured.
@@ -279,30 +289,34 @@ type Server struct {
 	// is the pre-R-82 shape.
 	tiers []BackupTier
 	// inFlight (R-85) is shared with the restore-test scheduler so the two never run together.
-	inFlight *backup.InFlight
+	inFlight           *backup.InFlight
 	afterPrimaryBackup func(ctx context.Context, vmid int) // the OS leg (agent v0.140.0); nil = none
-	logger   *slog.Logger
-	now      func() time.Time
+	lastKnown          LastKnownBackupStore                // R-894: on-disk newest success per tier; nil = none
+	logger             *slog.Logger
+	now                func() time.Time
 
-	disks           DiskOps             // slice 8C (optional)
-	diskGate        StorageGate         // slice 8C (optional)
-	guestList       GuestLister         // slice 8C (optional)
-	guestAttach     GuestAttacher       // slice 10 P2 (optional)
-	mem             MemoryOps           // v0.90.0 R-24 guest RAM resize (optional)
-	memMu           sync.Mutex          // single-flight around a resize apply (one customer per host)
-	netStorage      NetworkStorageOps   // Part A1: NAS network mounts (optional)
-	netMountRoot    string              // the user-data namespace root for the network-mount role gate
-	smbCredsDir     string              // where SMB creds files are written (out-of-band, 0600)
-	escrowStagePath string              // fork-4: 0600 staging file for the pushed restic repo password
+	disks           DiskOps           // slice 8C (optional)
+	diskGate        StorageGate       // slice 8C (optional)
+	guestList       GuestLister       // slice 8C (optional)
+	guestAttach     GuestAttacher     // slice 10 P2 (optional)
+	mem             MemoryOps         // v0.90.0 R-24 guest RAM resize (optional)
+	memMu           sync.Mutex        // single-flight around a resize apply (one customer per host)
+	netStorage      NetworkStorageOps // Part A1: NAS network mounts (optional)
+	netMountRoot    string            // the user-data namespace root for the network-mount role gate
+	smbCredsDir     string            // where SMB creds files are written (out-of-band, 0600)
+	escrowStagePath string            // fork-4: 0600 staging file for the pushed restic repo password
+	// crashGuardStatePath (R-856) is the host crash guard's state file read by GET /host/crash-guard;
+	// empty = defaultCrashGuardStatePath. A seam: tests point it at a fixture.
+	crashGuardStatePath string
 	// escrowRecovery (R-199, v0.125.0) assembles chain links 6-8: fetch this host's own sealed
 	// identity blob from the hub, unseal it with the customer's recovery code, return ONLY the
 	// offsite repository password. OPTIONAL — nil (no hub client configured) makes
 	// POST /escrow/recover-offsite-password answer 503 rather than pretending.
-	escrowRecovery  EscrowRecoverer
-	intent          IntentRecorder      // slice 10 P3 (optional)
-	guestBinds      *GuestBindStore     // F9 startup bind re-assert record (optional)
-	formatJobs      *FormatJobStore     // F20-BUG3 detached-format job record (optional)
-	staleLock       StaleLockController // F2-b startup stale-lock recovery (optional)
+	escrowRecovery EscrowRecoverer
+	intent         IntentRecorder      // slice 10 P3 (optional)
+	guestBinds     *GuestBindStore     // F9 startup bind re-assert record (optional)
+	formatJobs     *FormatJobStore     // F20-BUG3 detached-format job record (optional)
+	staleLock      StaleLockController // F2-b startup stale-lock recovery (optional)
 	// guestPower (F-REBOOT) is per-guest start-attempt state for the guest-power watchdog.
 	// Guarded by guestPowerMu in guestpower.go; in-memory on purpose (see guestPowerState).
 	guestPower map[int]guestPowerState
@@ -475,6 +489,7 @@ func NewServer(o Options) (*Server, error) {
 	s.tiers = normalizeBackupTiers(o.BackupTiers, o.Backups, cadence)
 	s.inFlight = o.InFlight
 	s.afterPrimaryBackup = o.AfterPrimaryBackup
+	s.lastKnown = o.LastKnownBackups
 	if s.backups == nil && len(s.tiers) > 0 {
 		s.backups = s.tiers[0].Service
 	}
@@ -520,6 +535,9 @@ func (s *Server) Handler() http.Handler {
 	// Host metrics (slice 9): host-wide health + per-storage capacity for the customer's monitoring
 	// view. Host-wide, token-authed, fresh (a live collect, not the 15-min hub snapshot).
 	mux.HandleFunc("GET /host/metrics", s.withGuest(s.handleHostMetrics))
+	// R-856 (`09` §3 decision 143): the host crash guard's record of the last HOST boot — the controller
+	// waits ~15 min with app mails after an unclean one. Read-only; a missing/garbled file = present:false.
+	mux.HandleFunc("GET /host/crash-guard", s.withGuest(s.handleCrashGuard))
 	// Disk management (slice 8C) — self-scoped; format routes through the data-bearing classifier+gate.
 	mux.HandleFunc("GET /disks", s.withGuest(s.handleDisks))
 	mux.HandleFunc("GET /disks/candidates", s.withGuest(s.handleDiskCandidates))
@@ -899,6 +917,13 @@ func (s *Server) handleBackup(w http.ResponseWriter, r *http.Request, vmid int)
 			s.logger.Info("local-api: backup job complete", "vmid", vmid, "target", tier.TargetID, "job", jobID, "archive", b.Archive)
 		}
 		s.store.RecordBackup(b)
+		if b.Success && s.lastKnown != nil {
+			if err := s.lastKnown.RecordBackupSuccess(tier.TargetID, b); err != nil {
+				// Not fatal: the backup exists. Only the fallback after a restart loses this copy.
+				s.logger.Warn("local-api: could not save the backup on disk for the due-check fallback (R-894)",
+					"vmid", vmid, "target", tier.TargetID, "err", err)
+			}
+		}
 		s.finishJob(key, jobID, b)
 		// OS leg (agent v0.140.0): after the night's whole-guest copy exists, still holding the heavy-op gate.
 		if b.Success && tier.Primary && s.afterPrimaryBackup != nil {
@@ -1061,6 +1086,20 @@ func (s *Server) handleBackupDue(w http.ResponseWriter, r *http.Request, vmid in
 		newest, haveNewest = t, true
 		unparseable = false // ground truth supersedes an unreadable in-memory timestamp
 	}
+	// R-894: the storage could not be read → the last success saved ON DISK stands in for the in-memory
+	// record a restart emptied. ONLY on archiveUnknown: a storage that answers is the ground truth, and an
+	// archive absent there must make the tier due even when the file remembers one (a pruned copy).
+	// A saved copy older than the cadence still reads DUE below — an unreadable storage never suppresses
+	// a backup that is due.
+	fromDisk := false
+	if lookup == archiveUnknown && s.lastKnown != nil {
+		if saved, ok := s.lastKnown.LastKnownSuccess(tier.TargetID, vmid); ok && (!haveNewest || saved.After(newest)) {
+			newest, haveNewest, fromDisk = saved, true, true
+			unparseable = false
+			s.logger.Info("local-api: backup storage unreadable — due-check uses the last success saved on disk (R-894)",
+				"vmid", vmid, "target", tier.TargetID, "saved", saved.UTC().Format(time.RFC3339))
+		}
+	}
 	if !haveNewest {
 		// R-88 Part 2: THREE distinct reasons for a nil age, each with its own state. Only ABSENT is a
 		// positive claim of "never backed up"; only that one may license the controller to bypass its
@@ -1083,13 +1122,17 @@ func (s *Server) handleBackupDue(w http.ResponseWriter, r *http.Request, vmid in
 	}
 	age := s.now().Sub(newest)
 	ageSecs := int64(age.Seconds())
+	suffix := ""
+	if fromDisk {
+		suffix = " (storage unreadable — age from the last success saved on disk)"
+	}
 	if age >= tier.Cadence {
 		writeOK(w, BackupDueResponse{VMID: vmid, Due: true, AgeSecs: &ageSecs, AgeState: AgeStateKnown,
-			Reason: "older than cadence", Target: echo})
+			Reason: "older than cadence" + suffix, Target: echo})
 		return
 	}
 	writeOK(w, BackupDueResponse{VMID: vmid, Due: false, AgeSecs: &ageSecs, AgeState: AgeStateKnown,
-		Reason: "within cadence window", Target: echo})
+		Reason: "within cadence window" + suffix, Target: echo})
 }
 
 // BackupTiersResponse is GET /backup/tiers (R-82): the tiers this agent serves, primary first.
@@ -1477,4 +1520,6 @@ func writeStatus(w http.ResponseWriter, code int, ok bool, data any, errMsg stri
 }
 
 // SetAfterPrimaryBackup wires the hook that runs after a successful primary-tier backup (the OS leg, agent v0.140.0).
-func (s *Server) SetAfterPrimaryBackup(fn func(ctx context.Context, vmid int)) { s.afterPrimaryBackup = fn }
+func (s *Server) SetAfterPrimaryBackup(fn func(ctx context.Context, vmid int)) {
+	s.afterPrimaryBackup = fn
+}

internal/localapi/tokenstore.go

@@ -154,12 +154,15 @@ func (s *TokenStore) Mint(vmid int) (string, error) {
 // looks it up; the per-candidate comparison is constant-time to avoid a timing oracle on the
 // stored hash. ok is false for an unknown/empty token.
 //
-// Reload-on-miss (B3): the store FILE is shared across processes — the one-shot provisioner
-// (`--selftest=provision`) Mints into it while the long-lived daemon serves Lookup from an index
-// built at open. On a miss, re-read the file ONCE and re-check, so a token minted after this
-// process started authorizes without a daemon restart (the drill's fresh-install 401). The
-// append-only log makes an unchanged file size proof of no new records, so a genuinely unknown
-// token costs at most one stat once the index is current — never a reload loop.
+// Reload-on-change (B3, R-269): the store FILE is shared across processes — the one-shot
+// provisioner (`--selftest=provision`) Mints into it while the long-lived daemon serves Lookup from
+// an index built at open. Every Lookup stats the file first and re-reads it when the append-only
+// log has grown, BEFORE answering — so a token minted elsewhere authorizes without a restart AND a
+// token rotated out elsewhere stops authorizing on its very next presentation. (Before R-269 the
+// re-read ran only on a MISS, so a superseded token was a direct map hit and kept authorizing until
+// some unrelated miss forced the reload.) An unchanged size is proof of no new records, so the
+// steady state costs one stat per call and never a reload loop. Pinned by
+// TestTokenStore_RotatedOutTokenRejectedFirst.
 func (s *TokenStore) Lookup(token string) (int, bool) {
 	if token == "" {
 		return 0, false
@@ -167,23 +170,34 @@ func (s *TokenStore) Lookup(token string) (int, bool) {
 	want := hashToken(token)
 	s.mu.Lock()
 	defer s.mu.Unlock()
-	// Direct map hit is the common path; the constant-time compare guards against a timing
-	// side-channel by re-checking the matched key (map lookup itself is not the secret-bearing
-	// comparison — the hash of a random 256-bit token is not feasibly guessable regardless).
-	if vmid, ok := s.byHash[want]; ok {
-		if subtle.ConstantTimeCompare([]byte(want), []byte(s.byVMID[vmid])) == 1 {
-			return vmid, true
+	st, statErr := os.Stat(s.path)
+	if statErr == nil && st.Size() != s.loadedSize {
+		// The log changed under us (another process minted/rotated): converge first, then answer.
+		s.reloads++
+		if err := s.reloadLocked(); err != nil {
+			return 0, false // unreadable store: fail closed, never crash the auth path
 		}
+		return s.matchLocked(want)
 	}
-	// Miss: skip the re-read when the append-only log has not grown (nothing new to see).
-	// A stat error falls through to the reload, which handles a missing file as empty.
-	if st, err := os.Stat(s.path); err == nil && st.Size() == s.loadedSize {
-		return 0, false
+	if vmid, ok := s.matchLocked(want); ok {
+		return vmid, true
 	}
+	if statErr == nil {
+		return 0, false // file unchanged since the last (re)load: genuinely unknown
+	}
+	// Stat failed (e.g. the file vanished): reload, which treats a missing file as empty.
 	s.reloads++
 	if err := s.reloadLocked(); err != nil {
-		return 0, false // unreadable store: fail closed, never crash the auth path
+		return 0, false
 	}
+	return s.matchLocked(want)
+}
+
+// matchLocked answers from the in-memory index. Direct map hit is the common path; the
+// constant-time compare re-checks the matched key against the guest's CURRENT hash (map lookup
+// itself is not the secret-bearing comparison — the hash of a random 256-bit token is not
+// feasibly guessable regardless). Caller holds the mutex.
+func (s *TokenStore) matchLocked(want string) (int, bool) {
 	if vmid, ok := s.byHash[want]; ok {
 		if subtle.ConstantTimeCompare([]byte(want), []byte(s.byVMID[vmid])) == 1 {
 			return vmid, true

internal/localapi/tokenstore_test.go

@@ -210,6 +210,51 @@ func TestTokenStore_ReloadOnMiss_RemintCoherence(t *testing.T) {
 	}
 }
 
+// R-269: a token rotated out by ANOTHER process must stop authorizing on its very next
+// presentation — with NO intervening lookup of the new token. This is the order the operator hits
+// after rotating a leaked token: the leaked one is presented first. RemintCoherence above looks the
+// NEW token up first, and that miss is what used to evict the old hash, so it passed while the leaked
+// token kept returning HTTP 200 on hardware (2026-08-09) until something unrelated forced a reload.
+//
+// RED-PROOF: restore the reload-on-MISS-only Lookup (answer a map hit before stat-ing the file) and
+// this fails with "rotated-out token still authorizes".
+func TestTokenStore_RotatedOutTokenRejectedFirst(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "tokens.log")
+	daemon, err := OpenTokenStore(path)
+	if err != nil {
+		t.Fatalf("open daemon store: %v", err)
+	}
+	defer daemon.Close()
+	minter, err := OpenTokenStore(path)
+	if err != nil {
+		t.Fatalf("open minter store: %v", err)
+	}
+	defer minter.Close()
+
+	old, err := minter.Mint(130)
+	if err != nil {
+		t.Fatalf("mint old: %v", err)
+	}
+	if vmid, ok := daemon.Lookup(old); !ok || vmid != 130 { // the daemon has learned the old token
+		t.Fatalf("old token before rotation: (%d,%v), want (130,true)", vmid, ok)
+	}
+	fresh, err := minter.Mint(130) // rotation, written by another process
+	if err != nil {
+		t.Fatalf("mint fresh: %v", err)
+	}
+
+	if vmid, ok := daemon.Lookup(old); ok { // the leaked token FIRST
+		t.Fatalf("rotated-out token still authorizes vmid %d on its first presentation after rotation — "+
+			"Mint's 'any previous token for this guest is revoked' is false across processes (R-269)", vmid)
+	}
+	if vmid, ok := daemon.Lookup(fresh); !ok || vmid != 130 {
+		t.Fatalf("fresh token after rotation: (%d,%v), want (130,true)", vmid, ok)
+	}
+	if vmid, ok := daemon.Lookup(old); ok {
+		t.Fatalf("rotated-out token authorizes vmid %d after the fresh one was seen", vmid)
+	}
+}
+
 // §8 edge: the store file deleted between open and a miss — reload treats it as empty; Lookup
 // fails closed, no crash.
 func TestTokenStore_ReloadOnMiss_MissingFile(t *testing.T) {

internal/osupdate/leg.go

@@ -106,6 +106,9 @@ type WrapperReport struct {
 	LiveRestore         json.RawMessage `json:"live_restore"`
 	Facts               json.RawMessage `json:"facts"`
 	Bundle              json.RawMessage `json:"bundle"` // the config bundle's result (R-840, mode "bundle")
+	// OOMCheck (R-528, `09` decision 157): the docker layer's memory-kill check, {result, oom_killed, oom_event,
+	// exit_code, image, detail}. Carried to the hub UNCHANGED; the agent never reads it.
+	OOMCheck json.RawMessage `json:"oom_check"`
 	// R-868 (v0.144.0): the agent's ids, echoed from the plan, so a report kept on disk can be sent without the
 	// agent process that started the pass. ReleaseID / VMID were always in the report.
 	RunID     string `json:"run_id"`
@@ -143,6 +146,9 @@ type Report struct {
 	DockerEngine        string          `json:"docker_engine,omitempty"` // docker layer: the engine after the step
 	Authority           string          `json:"authority,omitempty"`     // docker layer: ring0 | signed
 	Undo                bool            `json:"undo,omitempty"`          // docker layer: a signed undo (downgrade)
+	// OOMCheck: docker layer — the wrapper's oom_check object, byte-for-byte (R-528; the hub decides approval on it).
+	// Pinned by TestDocker_OOMCheckReachesTheHubUnchanged and TestR868_KeptCopyCarriesTheOOMCheck.
+	OOMCheck json.RawMessage `json:"oom_check,omitempty"`
 
 	unsent string // R-868: the wrapper's kept copy of this pass's report — deleted once the hub has it
 }
@@ -630,6 +636,7 @@ func (l *Leg) runLayer(ctx context.Context, runID, layer string, vmid int, trigg
 	}
 	rep.Upgraded, rep.PassSeconds = wr.Upgraded, wr.PassSeconds
 	rep.DockerEngine, rep.Authority, rep.Undo = wr.DockerEngine, wr.Authority, wr.Undo
+	rep.OOMCheck = rawOrNil(wr.OOMCheck)
 	if rep.Outcome == "" {
 		switch {
 		case rep.Mode == "inventory" && !blk.Enabled:
@@ -707,6 +714,14 @@ func (l *Leg) runLayer(ctx context.Context, runID, layer string, vmid int, trigg
 	return l.finish(ctx, lg, rep)
 }
 
+// rawOrNil: a wrapper field that is absent or JSON null stays out of the hub report (omitempty).
+func rawOrNil(m json.RawMessage) json.RawMessage {
+	if len(m) == 0 || string(m) == "null" {
+		return nil
+	}
+	return m
+}
+
 func onlyDocker(in []Package) []Package {
 	var out []Package
 	for _, p := range in {

internal/osupdate/leg_test.go

@@ -98,9 +98,13 @@ func (f *fakeWrapper) RunStdin(ctx context.Context, _ io.Reader, name string, ar
 	return f.Run(ctx, name, args...)
 }
 
-type fakeHub struct{ reports []Report }
+type fakeHub struct {
+	reports []Report
+	bodies  [][]byte // the exact bytes posted (R-528: the oom_check object must arrive unchanged)
+}
 
 func (h *fakeHub) PostOSReport(_ context.Context, body []byte) error {
+	h.bodies = append(h.bodies, append([]byte(nil), body...))
 	var r Report
 	json.Unmarshal(body, &r)
 	h.reports = append(h.reports, r)
@@ -364,7 +368,7 @@ func TestWrapperSuite(t *testing.T) {
 		t.Skip("python3 not available")
 	}
 	// the OS wrapper and (agent v0.142.0) the crash guard — both root programs in configs/ with their own suites
-	for _, suite := range []string{"../../configs/test_felhom_os_apply.py", "../../configs/test_felhom_crash_guard.py", "../../configs/test_felhom_config_bundle.py"} {
+	for _, suite := range []string{"../../configs/test_felhom_os_apply.py", "../../configs/test_felhom_crash_guard.py", "../../configs/test_felhom_config_bundle.py", "../../configs/test_felhom_priv_apply.py"} {
 		cmd := exec.Command(py, "-B", suite)
 		out, err := cmd.CombinedOutput()
 		if err != nil {
@@ -504,3 +508,42 @@ func TestHealthVerdict_ControllerBlindToDockerFails(t *testing.T) {
 		t.Fatal("an older wrapper (no field) must not fail")
 	}
 }
+
+// R-528 (`09` decision 157): the wrapper's oom_check object reaches the hub's docker report byte-for-byte; the guest
+// and host reports carry none. COMPANION RED-PROOF: drop `rep.OOMCheck = rawOrNil(wr.OOMCheck)` in runLayer → "no
+// oom_check in the docker report".
+const oomCheckWire = `{"detail":"the engine reported the memory kill: OOMKilled=true and the oom event","exit_code":137,"image":"gitea.dooplex.hu/admin/felhom-controller:0.300.0","oom_event":true,"oom_killed":true,"result":"pass"}`
+
+func TestDocker_OOMCheckReachesTheHubUnchanged(t *testing.T) {
+	w := &fakeWrapper{t: t, applyRep: map[string]WrapperReport{LayerDocker: {
+		Upgraded:     []Package{{Name: "docker-ce", Version: "5:29.8.2-1~debian.13~trixie"}},
+		DockerEngine: "29.8.2", Authority: "ring0", OOMCheck: json.RawMessage(oomCheckWire)}}}
+	l, h := newLeg(t, w, &hub.WireOSUpdate{Ring: 0, Enabled: true})
+	l.Run(context.Background(), 9201, "night")
+	found := false
+	for _, b := range h.bodies {
+		var m map[string]json.RawMessage
+		if err := json.Unmarshal(b, &m); err != nil {
+			t.Fatal(err)
+		}
+		var layer string
+		json.Unmarshal(m["layer"], &layer)
+		oc, has := m["oom_check"]
+		if layer != LayerDocker {
+			if has {
+				t.Fatalf("the %s report carries an oom_check: %s", layer, oc)
+			}
+			continue
+		}
+		found = true
+		if !has {
+			t.Fatalf("no oom_check in the docker report: %s", b)
+		}
+		if string(oc) != oomCheckWire {
+			t.Fatalf("oom_check changed on the way:\n got %s\nwant %s", oc, oomCheckWire)
+		}
+	}
+	if !found {
+		t.Fatalf("no docker report posted: %s", calls(w))
+	}
+}

internal/osupdate/unsent.go

@@ -7,6 +7,7 @@ import (
 	"path/filepath"
 	"strings"
 	"syscall"
+	"time"
 
 	"gitea.dooplex.hu/admin/felhom-agent/internal/hub"
 )
@@ -62,6 +63,24 @@ func (l *Leg) SendUnsent(ctx context.Context) int {
 	return l.sendUnsentLocked(ctx)
 }
 
+// SendUnsentLoop runs SendUnsent now and then every `every` until ctx ends (v0.144.1). MEASURED live on demo-hp
+// 2026-10-05: after a kill -9 the daemon restarted in ~5 s, while the orphaned root wrapper was still installing — its
+// copy appeared ~7 s AFTER the start-time sender had looked. One look at start is therefore not enough. A glob of the
+// plan dir every few minutes costs nothing; the pass lock keeps it off a running pass. Pinned by
+// TestR868_ACopyWrittenAfterTheStartIsSentByTheLoop.
+func (l *Leg) SendUnsentLoop(ctx context.Context, every time.Duration, onSent func(int)) {
+	for {
+		if n := l.SendUnsent(ctx); n > 0 && onSent != nil {
+			onSent(n)
+		}
+		select {
+		case <-ctx.Done():
+			return
+		case <-time.After(every):
+		}
+	}
+}
+
 func (l *Leg) sendUnsentLocked(ctx context.Context) int {
 	if l.Hub == nil {
 		return 0 // nobody to send to: keep the copies for a process that has the hub
@@ -82,7 +101,7 @@ func (l *Leg) sendUnsentLocked(ctx context.Context) int {
 		}
 		rep := l.reportFromKept(ctx, wr, f)
 		lg := l.log().With("run", rep.RunID, "layer", rep.Layer, "vmid", rep.VMID, "ring", rep.Ring, "trigger", rep.Trigger)
-		lg.Info("osupdate: sending a report the agent never sent (the agent stopped mid-pass, R-868)", "path", f)
+		lg.Info("osupdate: sending a kept report late (the agent was stopped mid-pass or the hub was away, R-868)", "path", f)
 		before := rep.unsent
 		_ = l.finish(ctx, lg, rep)
 		if _, err := os.Stat(before); os.IsNotExist(err) {
@@ -111,7 +130,7 @@ func (l *Leg) reportFromKept(ctx context.Context, wr WrapperReport, path string)
 	}
 	rep := Report{RunID: runID, Layer: wr.Layer, Trigger: trigger, Mode: wr.Mode, Ring: ring, ReleaseID: wr.ReleaseID,
 		VMID: wr.VMID, unsent: path}
-	prefix := "sent after the agent stopped mid-pass (R-868)"
+	prefix := "sent late — kept on the box until the hub could take it (R-868, R-875)"
 	switch {
 	case wr.refused():
 		rep.Outcome, rep.Refused, rep.HealthReason = "refused", wr.Refused, prefix
@@ -125,6 +144,7 @@ func (l *Leg) reportFromKept(ctx context.Context, wr WrapperReport, path string)
 	}
 	rep.Upgraded, rep.PassSeconds = wr.Upgraded, wr.PassSeconds
 	rep.DockerEngine, rep.Authority, rep.Undo = wr.DockerEngine, wr.Authority, wr.Undo
+	rep.OOMCheck = rawOrNil(wr.OOMCheck)
 	wantEngine := ""
 	for _, u := range wr.Upgraded {
 		if u.Name == "docker-ce" {

internal/osupdate/unsent_test.go

@@ -6,7 +6,9 @@ import (
 	"errors"
 	"os"
 	"path/filepath"
+	"strings"
 	"testing"
+	"time"
 
 	"gitea.dooplex.hu/admin/felhom-agent/internal/hub"
 )
@@ -37,6 +39,10 @@ func TestR868_KilledPassIsReportedAtStart(t *testing.T) {
 	if r.Outcome != "applied" || !r.Healthy || r.Trigger != "debug" || r.RunID != kept.RunID || r.Ring != 0 || len(r.Upgraded) != 2 || r.VMID != 9201 {
 		t.Fatalf("report = %+v", r)
 	}
+	// R-875 (v0.145.0): neutral — the copy cannot tell a killed agent from an absent hub.
+	if !strings.HasPrefix(r.HealthReason, "sent late") || strings.Contains(r.HealthReason, "stopped mid-pass") {
+		t.Fatalf("health_reason = %q, want the neutral \"sent late …\"", r.HealthReason)
+	}
 	for _, p := range []string{rp, pp} {
 		if _, err := os.Stat(p); !os.IsNotExist(err) {
 			t.Fatalf("%s still on disk after the hub got it", filepath.Base(p))
@@ -115,9 +121,53 @@ func TestR868_RunningPassKeepsTheSenderOff(t *testing.T) {
 	}
 }
 
+// v0.144.1 — THE LIVE SHAPE (demo-hp 2026-10-05 05:45 UTC): the restarted daemon looked at 05:45:09, the orphaned
+// wrapper wrote its copy at ~05:45:16. The loop must still send it.
+// COMPANION RED-PROOF: make SendUnsentLoop return after the first look → "the late copy was never sent".
+func TestR868_ACopyWrittenAfterTheStartIsSentByTheLoop(t *testing.T) {
+	w := &fakeWrapper{t: t}
+	l, h := newLeg(t, w, nil)
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	sent := make(chan int, 4)
+	go l.SendUnsentLoop(ctx, 20*time.Millisecond, func(n int) { sent <- n })
+	time.Sleep(50 * time.Millisecond) // the start-time look found nothing
+	must(t, os.WriteFile(reportFile(l.PlanDir, "late", LayerGuest, "apply"),
+		[]byte(`{"mode":"apply","layer":"guest","run_id":"late","trigger":"debug","ring":0,"vmid":9201,"upgraded":[{"name":"openssl","version":"u3"}]}`), 0o600))
+	select {
+	case n := <-sent:
+		if n != 1 || len(h.reports) != 1 || h.reports[0].RunID != "late" || h.reports[0].Outcome == "" {
+			t.Fatalf("sent %d: %+v", n, h.reports)
+		}
+	case <-time.After(3 * time.Second):
+		t.Fatal("the late copy was never sent")
+	}
+}
+
 func must(t *testing.T, err error) {
 	t.Helper()
 	if err != nil {
 		t.Fatal(err)
 	}
 }
+
+// R-528: a kept docker report (the agent was killed) still carries the oom_check object to the hub unchanged.
+// COMPANION RED-PROOF: drop `rep.OOMCheck = rawOrNil(wr.OOMCheck)` in reportFromKept → "oom_check lost".
+func TestR868_KeptCopyCarriesTheOOMCheck(t *testing.T) {
+	w := &fakeWrapper{t: t}
+	l, h := newLeg(t, w, &hub.WireOSUpdate{Ring: 0, Enabled: true})
+	ring := 0
+	kept := WrapperReport{Mode: "apply", Layer: LayerDocker, RunID: "20261007T020000Z", Trigger: "night", Ring: &ring, VMID: 9201,
+		ReleaseID: "ring0-20261007T020000Z", HealthBefore: guestOK(), HealthAfter: guestOK(), DockerEngine: "29.8.2",
+		Upgraded: []Package{{Name: "docker-ce", Version: "5:29.8.2-1~debian.13~trixie"}}, OOMCheck: json.RawMessage(oomCheckWire)}
+	b, _ := json.Marshal(kept)
+	must(t, os.WriteFile(reportFile(l.PlanDir, kept.RunID, LayerDocker, "apply"), b, 0o600))
+	if n := l.SendUnsent(context.Background()); n != 1 || len(h.bodies) != 1 {
+		t.Fatalf("sent %d, bodies %d", n, len(h.bodies))
+	}
+	var m map[string]json.RawMessage
+	must(t, json.Unmarshal(h.bodies[0], &m))
+	if string(m["oom_check"]) != oomCheckWire {
+		t.Fatalf("oom_check lost or changed: %q", m["oom_check"])
+	}
+}

internal/pbs/live_reporter_test.go

@@ -60,8 +60,8 @@ func TestLiveReporter_CoordPresentWithoutPriorVerify(t *testing.T) {
 	if h.PBS == nil {
 		t.Fatal("pbs coord absent despite a reachable PBS — the gap this fixes")
 	}
-	if h.PBS.RepoID != "felhom-pbs" || h.PBS.Namespace != "root" || h.PBS.LatestSnapshotID != "9201" {
-		t.Errorf("pbs coord = %+v, want felhom-pbs/root/9201", h.PBS)
+	if h.PBS.RepoID != "felhom-pbs" || h.PBS.Namespace != hub.PBSRootNamespace || h.PBS.LatestSnapshotID != "9201" {
+		t.Errorf("pbs coord = %+v, want felhom-pbs, the root namespace (\"\", R-124), 9201", h.PBS)
 	}
 
 	// COMPANION (pre-fix): the bare SnapshotStore (no live read) with an empty store omits pbs.

internal/privapplytest/check.go

@@ -0,0 +1,34 @@
+// Package privapplytest runs configs/felhom-priv-apply in CHECK-ONLY mode from Go tests (R-861): each renderer's
+// real output must be accepted by the root checker, so the two can never drift apart unnoticed. Test-only helper.
+package privapplytest
+
+import (
+	"os"
+	"os/exec"
+	"path/filepath"
+	"runtime"
+	"strings"
+	"testing"
+)
+
+// Check writes content to a temp file and runs `felhom-priv-apply --check <verb> [name] <file>`. It returns the
+// checker's verdict line ("OK" or "REFUSED [rule] …"). Skips when python3 is absent.
+func Check(t *testing.T, verb, name, content string) string {
+	t.Helper()
+	py, err := exec.LookPath("python3")
+	if err != nil {
+		t.Skip("python3 not available")
+	}
+	_, here, _, _ := runtime.Caller(0)
+	wrapper := filepath.Join(filepath.Dir(here), "..", "..", "configs", "felhom-priv-apply")
+	f := filepath.Join(t.TempDir(), "staged")
+	if err := os.WriteFile(f, []byte(content), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	args := []string{"-B", wrapper, "--check", verb}
+	if name != "" {
+		args = append(args, name)
+	}
+	out, _ := exec.Command(py, append(args, f)...).CombinedOutput()
+	return strings.TrimSpace(string(out))
+}

internal/provision/backhalf.go

@@ -186,8 +186,9 @@ func (b *BackHalf) Provision(ctx context.Context, in Input) (Result, error) {
 	// 6. Install + register the pre-start self-heal hook (C1 net): if a data drive is absent at a future
 	//    boot, the hook creates a placeholder for its missing bind source so the guest still starts.
 	//    Best-effort + non-fatal — it's defense-in-depth; a provision must not fail over the hook.
-	if err := guesthook.InstallSnippet(ctx, b.runner); err != nil {
-		b.logger.Warn("provision: pre-start hook snippet install failed (non-fatal)", "vmid", in.VMID, "err", err)
+	//    R-861 (v0.146.0): the hook FILE comes with the signed config bundle; the agent only checks it and registers it.
+	if err := guesthook.SnippetReady(guesthook.SnippetPath); err != nil {
+		b.logger.Warn("provision: pre-start hook not in place — not registering it (non-fatal)", "vmid", in.VMID, "err", err)
 	} else if err := guesthook.Register(ctx, b.runner, in.VMID); err != nil {
 		b.logger.Warn("provision: pre-start hook registration failed (non-fatal)", "vmid", in.VMID, "err", err)
 	}

internal/selfupdate/executor.go

@@ -3,6 +3,7 @@ package selfupdate
 import (
 	"context"
 	"crypto/sha256"
+	"encoding/base64"
 	"encoding/hex"
 	"encoding/json"
 	"fmt"
@@ -11,6 +12,7 @@ import (
 	"net/http"
 	"os"
 	"path/filepath"
+	"strings"
 	"time"
 
 	"gitea.dooplex.hu/admin/felhom-agent/internal/signedjobs"
@@ -47,6 +49,7 @@ type Executor struct {
 	username    string
 	token       string
 	stateDir    string
+	planDir     string // felhom-os-apply's plan dir; "" = defaultPlanDir (tests set a temp dir)
 	runner      WrapperRunner
 	httpClient  *http.Client
 	logger      *slog.Logger
@@ -126,18 +129,70 @@ func (e *Executor) Execute(ctx context.Context, op string, params json.RawMessag
 		return fmt.Errorf("agent_update: chmod staged: %w", err)
 	}
 
-	// Hand off to the root wrapper. It re-verifies the sha, flips A/B, writes the pending marker,
-	// and schedules the detached restart. After this the new binary starts; the commit is the
-	// Manager's job once it has dwelled cleanly.
-	e.logger.Warn("agent_update: handing staged binary to the guarded wrapper", "staged", staged, "version", p.Version)
-	stdout, stderr, err := e.runner.Run(ctx, wrapperPath, "apply", staged, p.SHA256)
-	if err != nil {
-		return fmt.Errorf("agent_update: wrapper apply failed: %w (stderr: %s)", err, string(stderr))
+	// R-861 (v0.146.0): hand the SIGNED job to the root wrapper felhom-os-apply (mode agent_update). It verifies the
+	// operator's signature itself (root-owned signers file, this host, the window, the nonce), re-hashes the staged
+	// file against the SIGNED sha, and only then runs the A/B flip (felhom-selfupdate-guarded apply, no longer in the
+	// agent's sudoers). Until v0.146.0 the agent passed the sha to the flip itself, so a compromised agent could
+	// install any binary — and the binary is what the escrow ceremony and the guest hook run as root.
+	so, ok := signedjobs.SignedOpFrom(ctx)
+	if !ok {
+		return fmt.Errorf("agent_update: no signed envelope in the context — the root wrapper could not verify it")
 	}
-	e.logger.Warn("agent_update: apply handed off; restart scheduled", "version", p.Version, "wrapper", trim(stdout))
+	planDir := e.planDir
+	if planDir == "" {
+		planDir = defaultPlanDir
+	}
+	if err := os.MkdirAll(planDir, 0o700); err != nil {
+		return fmt.Errorf("agent_update: plan dir: %w", err)
+	}
+	plan, _ := json.Marshal(map[string]any{"release_id": "agent-" + p.Version, "layer": "host", "mode": "agent_update",
+		"staged": staged, "signed": map[string]string{"blob_b64": base64.StdEncoding.EncodeToString(so.Blob), "sig": string(so.Sig)}})
+	planPath := filepath.Join(planDir, "plan-agentupdate-"+p.Version+".json")
+	if err := os.WriteFile(planPath, plan, 0o600); err != nil {
+		return fmt.Errorf("agent_update: write plan: %w", err)
+	}
+	defer os.Remove(planPath)
+	e.logger.Warn("agent_update: handing the signed job to the root wrapper (felhom-os-apply agent_update)", "staged", staged, "version", p.Version)
+	stdout, stderr, err := e.runner.Run(ctx, osApplyPath, "--plan", planPath)
+	rep := parseOSApplyReport(stdout)
+	var r struct {
+		Refused     json.RawMessage `json:"refused"`
+		Failed      json.RawMessage `json:"failed"`
+		AgentUpdate json.RawMessage `json:"agent_update"`
+	}
+	jerr := json.Unmarshal([]byte(rep), &r)
+	refused := len(r.Refused) > 0 && string(r.Refused) != "null"
+	if err != nil || jerr != nil || refused || len(r.Failed) > 0 || len(r.AgentUpdate) == 0 {
+		return fmt.Errorf("agent_update: the root wrapper did not apply it: %v (report: %s; stderr: %s)", err, trimStr(rep), trim(stderr))
+	}
+	e.logger.Warn("agent_update: signed update verified as root and handed off; restart scheduled", "version", p.Version, "report", trimStr(rep))
 	return nil
 }
 
+// osApplyPath is the root wrapper that verifies the signed agent_update (R-861). Fixed, never config-overridable.
+const osApplyPath = "/usr/local/sbin/felhom-os-apply"
+
+// defaultPlanDir is felhom-os-apply's ONLY plan directory (PLAN_DIR there; osupdate.DefaultPlanDir here).
+const defaultPlanDir = "/var/lib/felhom-agent/os"
+
+// parseOSApplyReport returns the JSON after the wrapper's last "OSAPPLY-REPORT " line ("" when there is none).
+func parseOSApplyReport(stdout []byte) string {
+	rep := ""
+	for _, line := range strings.Split(string(stdout), "\n") {
+		if strings.HasPrefix(line, "OSAPPLY-REPORT ") {
+			rep = strings.TrimPrefix(line, "OSAPPLY-REPORT ")
+		}
+	}
+	return rep
+}
+
+func trimStr(s string) string {
+	if len(s) > 400 {
+		return s[:400] + "…"
+	}
+	return s
+}
+
 // download streams url → dest (0644, fsync'd) and returns the lowercase-hex sha256 of the bytes.
 func (e *Executor) download(ctx context.Context, url, dest string) (string, error) {
 	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)

internal/selfupdate/executor_test.go

@@ -15,34 +15,56 @@ import (
 	"sync"
 	"testing"
 
+	"gitea.dooplex.hu/admin/felhom-agent/internal/reconcile"
 	"gitea.dooplex.hu/admin/felhom-agent/internal/signedjobs"
 )
 
-// fakeWrapper records the verbs the executor/manager shell out, and returns a configurable error.
+// fakeWrapper records the verbs the executor/manager shell out, and returns a configurable error. For the os-apply
+// call it snapshots the plan file at call time (the executor removes it afterwards) and answers with report.
 type fakeWrapper struct {
-	mu    sync.Mutex
-	calls [][]string
-	err   error
+	mu     sync.Mutex
+	calls  [][]string
+	err    error
+	plans  []map[string]any
+	report string // the OSAPPLY-REPORT JSON; "" = a successful agent_update
 }
 
 func (f *fakeWrapper) Run(_ context.Context, name string, args ...string) ([]byte, []byte, error) {
 	f.mu.Lock()
 	defer f.mu.Unlock()
 	f.calls = append(f.calls, append([]string{name}, args...))
+	if name == osApplyPath && len(args) == 2 && args[0] == "--plan" {
+		var p map[string]any
+		b, _ := os.ReadFile(args[1])
+		_ = json.Unmarshal(b, &p)
+		f.plans = append(f.plans, p)
+		rep := f.report
+		if rep == "" {
+			rep = `{"agent_update": {"version": "x", "wrapper_rc": 0}, "mode": "agent_update", "refused": null}`
+		}
+		return []byte("OSAPPLY-REPORT " + rep + "\n"), nil, f.err
+	}
 	return []byte("ok"), nil, f.err
 }
+
+// applyCalls are the hand-offs to the root wrapper (felhom-os-apply --plan …). Since v0.146.0 the agent never calls
+// `felhom-selfupdate-guarded apply` itself.
 func (f *fakeWrapper) applyCalls() [][]string {
 	f.mu.Lock()
 	defer f.mu.Unlock()
 	var out [][]string
 	for _, c := range f.calls {
-		if len(c) >= 2 && c[1] == "apply" {
+		if c[0] == osApplyPath || (len(c) >= 2 && c[1] == "apply") {
 			out = append(out, c)
 		}
 	}
 	return out
 }
 
+func signedCtx() context.Context {
+	return signedjobs.WithSignedOp(context.Background(), &reconcile.SignedOp{Blob: []byte(`{"op":"agent_update"}`), Sig: []byte("SIG")})
+}
+
 func sha256Of(b []byte) string {
 	h := sha256.Sum256(b)
 	return hex.EncodeToString(h[:])
@@ -65,6 +87,7 @@ func newExec(t *testing.T, srv *httptest.Server, wrap WrapperRunner) (*Executor,
 		Runner:      wrap,
 		Logger:      slog.New(slog.NewTextHandler(io.Discard, nil)),
 	})
+	e.planDir = t.TempDir()
 	return e, stateDir
 }
 
@@ -84,7 +107,7 @@ func TestExecutor_HappyPath(t *testing.T) {
 	e, stateDir := newExec(t, srv, wrap)
 
 	sha := sha256Of(body)
-	if err := e.Execute(context.Background(), "agent_update", updateParamsJSON(t, "0.70.1", sha)); err != nil {
+	if err := e.Execute(signedCtx(), "agent_update", updateParamsJSON(t, "0.70.1", sha)); err != nil {
 		t.Fatalf("execute: %v", err)
 	}
 	staged := filepath.Join(stateDir, "selfupdate", "felhom-agent-0.70.1")
@@ -93,11 +116,46 @@ func TestExecutor_HappyPath(t *testing.T) {
 		t.Fatalf("staged binary missing/mismatch: %v", err)
 	}
 	calls := wrap.applyCalls()
-	if len(calls) != 1 {
-		t.Fatalf("apply invoked %d times, want 1 (%v)", len(calls), wrap.calls)
+	if len(calls) != 1 || calls[0][0] != osApplyPath || calls[0][1] != "--plan" {
+		t.Fatalf("want exactly one hand-off to felhom-os-apply --plan, got %v", wrap.calls)
 	}
-	if calls[0][2] != staged || calls[0][3] != sha {
-		t.Errorf("apply args = %v, want [.. apply %s %s]", calls[0], staged, sha)
+	// R-861: the plan carries the SIGNED envelope and the staged path; the wrapper, not the agent, decides.
+	p := wrap.plans[0]
+	sg, _ := p["signed"].(map[string]any)
+	if p["mode"] != "agent_update" || p["layer"] != "host" || p["staged"] != staged || sg["sig"] != "SIG" || sg["blob_b64"] == "" {
+		t.Errorf("plan = %v, want mode agent_update + the staged path + the signed envelope", p)
+	}
+	if _, err := os.Stat(calls[0][2]); !os.IsNotExist(err) {
+		t.Error("the plan file was left behind")
+	}
+}
+
+// R-861: without the signed envelope nothing reaches the root wrapper.
+// RED-PROOF (audits/hub-safety-2026-10-05/partF/red-proof.txt): call `felhom-selfupdate-guarded apply` directly again
+// → the happy path's "exactly one hand-off to felhom-os-apply" fails.
+func TestExecutor_NoEnvelopeNoHandOff(t *testing.T) {
+	body := []byte("good bytes")
+	srv := artifactServer(t, body)
+	defer srv.Close()
+	wrap := &fakeWrapper{}
+	e, _ := newExec(t, srv, wrap)
+	if err := e.Execute(context.Background(), "agent_update", updateParamsJSON(t, "0.70.1", sha256Of(body))); err == nil {
+		t.Fatal("an update with no signed envelope was handed on")
+	}
+	if len(wrap.applyCalls()) != 0 {
+		t.Fatalf("the root wrapper was called without an envelope: %v", wrap.calls)
+	}
+}
+
+// A refusal in the wrapper's report is a failure, even when its exit code reads 0.
+func TestExecutor_WrapperRefusalSurfaces(t *testing.T) {
+	body := []byte("good bytes")
+	srv := artifactServer(t, body)
+	defer srv.Close()
+	wrap := &fakeWrapper{report: `{"mode": "agent_update", "refused": {"code": "R3", "reason": "the operator signature does not verify"}}`}
+	e, _ := newExec(t, srv, wrap)
+	if err := e.Execute(signedCtx(), "agent_update", updateParamsJSON(t, "0.70.1", sha256Of(body))); err == nil {
+		t.Fatal("a refused signed update read as applied")
 	}
 }
 
@@ -158,7 +216,7 @@ func TestExecutor_WrapperFailureSurfaces(t *testing.T) {
 	defer srv.Close()
 	wrap := &fakeWrapper{err: errors.New("wrapper refused: sha mismatch")}
 	e, _ := newExec(t, srv, wrap)
-	if err := e.Execute(context.Background(), "agent_update", updateParamsJSON(t, "0.70.1", sha256Of(body))); err == nil {
+	if err := e.Execute(signedCtx(), "agent_update", updateParamsJSON(t, "0.70.1", sha256Of(body))); err == nil {
 		t.Fatal("wrapper failure must surface")
 	}
 }

internal/storage/hostops.go

@@ -57,6 +57,10 @@ type DeviceProbe struct {
 	HasPartitions     bool   `json:"has_partitions"`      // child partitions present (lsblk)
 	Mounted           bool   `json:"mounted"`             // currently mounted somewhere
 	FSType            string `json:"fstype,omitempty"`
+	// FSUUID is the filesystem UUID blkid read from the on-disk superblock (`blkid -p`, no cache),
+	// "" when there is none. R-25: the format path reports it so the caller mounts the filesystem the
+	// agent just made, not whatever a /dev path resolves to later.
+	FSUUID string `json:"fs_uuid,omitempty"`
 }
 
 // DataBearing is the conservative verdict: any signature / partition table / partition / mount —
@@ -101,6 +105,10 @@ type MountSpec struct {
 type Binaries struct {
 	Systemctl   string
 	Install     string
+	// PrivApply is the root content checker (R-861, agent v0.146.0): a unit file reaches /etc/systemd/system only
+	// through `felhom-priv-apply unit <name>`, which reads the staged copy from /var/lib/felhom-agent/units/ and
+	// refuses anything the renderers here never produce (a bind over /etc, a Where outside /mnt, …).
+	PrivApply string
 	Smartctl    string
 	Lvs         string
 	Blkid       string // device signature probe (8C data-bearing detection)
@@ -119,6 +127,9 @@ func (b Binaries) withDefaults() Binaries {
 	if b.Install == "" {
 		b.Install = "/usr/bin/install"
 	}
+	if b.PrivApply == "" {
+		b.PrivApply = "/usr/local/sbin/felhom-priv-apply"
+	}
 	if b.Smartctl == "" {
 		b.Smartctl = "/usr/sbin/smartctl"
 	}
@@ -246,9 +257,9 @@ func (h *SudoHostOps) EnsureMount(ctx context.Context, spec MountSpec) error {
 		return fmt.Errorf("storage: staging unit: %w", err)
 	}
 
-	dest := filepath.Join(h.unitDir, unitName)
-	// install as root (atomic copy with fixed mode/owner) — fixed arg vector.
-	if err := h.run(ctx, h.bins.Install, "-o", "root", "-g", "root", "-m", "0644", "--", stagePath, dest); err != nil {
+	// R-861: the root checker installs it (fixed source dir, fixed destination dir, content checked) — never a
+	// plain `install` of a file the agent wrote.
+	if err := h.run(ctx, h.bins.PrivApply, "unit", unitName); err != nil {
 		return fmt.Errorf("storage: installing unit %s: %w", unitName, err)
 	}
 	if err := h.run(ctx, h.bins.Systemctl, "daemon-reload"); err != nil {
@@ -416,6 +427,8 @@ func (h *SudoHostOps) InspectDevice(ctx context.Context, device string) (DeviceP
 			probe.FSType = v
 		case "PTTYPE":
 			probe.HasPartitionTable = true
+		case "UUID":
+			probe.FSUUID = v
 		case "USAGE":
 			if v != "" {
 				probe.HasFilesystem = true // filesystem/raid/crypto member = data-bearing

internal/storage/hostops_disk_test.go

@@ -208,3 +208,17 @@ func TestFormat_RejectsBadArgs(t *testing.T) {
 		t.Fatalf("mkfs ran despite invalid input: %v", r.calls)
 	}
 }
+
+// R-25: the probe carries the superblock's filesystem UUID so the format path can report the new one.
+func TestInspect_ReadsFilesystemUUID(t *testing.T) {
+	r := &scriptedRunner{
+		outputs: map[string][]byte{
+			"blkid": []byte("DEVNAME=/dev/sdb\nUUID=0fc63daf-8483-4772-8e79-3d69d8477de4\nTYPE=ext4\nUSAGE=filesystem\n"),
+			"lsblk": []byte(`{"blockdevices":[{"name":"sdb","fstype":"ext4","pttype":null,"mountpoint":null}]}`),
+		},
+	}
+	p, _ := newSudo(r).InspectDevice(context.Background(), "/dev/sdb")
+	if p.FSUUID != "0fc63daf-8483-4772-8e79-3d69d8477de4" {
+		t.Fatalf("FSUUID = %q, want the blkid UUID", p.FSUUID)
+	}
+}

internal/storage/hostops_test.go

@@ -55,8 +55,9 @@ func TestHostOps_MountLifecycle(t *testing.T) {
 	if len(rr.calls) != 3 {
 		t.Fatalf("expected 3 commands, got %d: %v", len(rr.calls), rr.calls)
 	}
-	if rr.calls[0][0] != "/usr/bin/install" || !contains(rr.calls[0], "0644") {
-		t.Errorf("call[0] not the install: %v", rr.calls[0])
+	// R-861: the unit is installed by the root content checker, named — never a plain `install` of a staged path.
+	if rr.calls[0][0] != "/usr/local/sbin/felhom-priv-apply" || len(rr.calls[0]) != 3 || rr.calls[0][1] != "unit" {
+		t.Errorf("call[0] not the checker's unit install: %v", rr.calls[0])
 	}
 	if !contains(rr.calls[1], "daemon-reload") {
 		t.Errorf("call[1] not daemon-reload: %v", rr.calls[1])

internal/storage/netmigrate_test.go

@@ -100,7 +100,7 @@ func TestMigrateNetworkUnits_RewritesDriftedOnceIdempotent(t *testing.T) {
 		if strings.Contains(joined, "daemon-reload") {
 			reloads++
 		}
-		if strings.Contains(joined, "install") {
+		if strings.Contains(joined, "felhom-priv-apply unit") {
 			installs++
 		}
 	}

internal/storage/netmount.go

@@ -236,9 +236,13 @@ func (s NetworkMountSpec) mountOptions() string {
 			"file_mode=0664",
 			"dir_mode=0775",
 			"_netdev",
+			"nosuid",
+			"nodev",
 		}, ",")
 	}
-	return "vers=4.1,soft,timeo=50,retrans=2,noatime,_netdev,retry=0"
+	// R-861 (v0.146.0): nosuid,nodev — a set-uid file or a device node on a server outside the box must never act on
+	// the host. felhom-priv-apply refuses a network unit without them.
+	return "vers=4.1,soft,timeo=50,retrans=2,noatime,_netdev,retry=0,nosuid,nodev"
 }
 
 // renderNetworkMountUnit builds the .mount unit (triggered by the .automount; deliberately NO [Install]
@@ -409,8 +413,8 @@ func (h *SudoHostOps) installUnit(ctx context.Context, unitName, content string)
 	if err := os.WriteFile(stagePath, []byte(content), 0o644); err != nil {
 		return fmt.Errorf("netmount: staging unit %s: %w", unitName, err)
 	}
-	dest := filepath.Join(h.unitDir, unitName)
-	if err := h.run(ctx, h.bins.Install, "-o", "root", "-g", "root", "-m", "0644", "--", stagePath, dest); err != nil {
+	// R-861: through the root checker (felhom-priv-apply unit), never a plain install of an agent-written file.
+	if err := h.run(ctx, h.bins.PrivApply, "unit", unitName); err != nil {
 		return fmt.Errorf("netmount: installing unit %s: %w", unitName, err)
 	}
 	return nil

internal/storage/netmount_test.go

@@ -264,7 +264,7 @@ func TestEnsureNetworkMount_Commands(t *testing.T) {
 		switch {
 		case strings.Contains(joined, "mkdir") && strings.Contains(joined, "/mnt/felhom-drives/media"):
 			sawMkdir = true
-		case strings.Contains(joined, "install"):
+		case strings.Contains(joined, "felhom-priv-apply unit"):
 			installs++
 		case strings.Contains(joined, "daemon-reload"):
 			sawReload = true

internal/storage/r330_smart_attrs_test.go

@@ -0,0 +1,64 @@
+package storage
+
+import (
+	"encoding/json"
+	"strings"
+	"testing"
+)
+
+// R-330 (disk health Phase 2) — attributes 187, 188 and 199 ride the wire.
+//
+// The failing drive of 2026-08-14 carried 187 Reported_Uncorrect at raw 1001 while SMART said PASSED;
+// none of the three reached the controller. The values are carried as RAW counters; an absent
+// attribute is OMITTED from the JSON (unknown), never sent as 0 (S-39).
+
+// The 2026-08-14 shape: PASSED, 187 raw 1001, plus 188/199 and the existing three.
+const r330SATA = `{"smart_status":{"passed":true},"ata_smart_attributes":{"table":[
+ {"id":5,"raw":{"value":0}},
+ {"id":187,"raw":{"value":1001}},
+ {"id":188,"raw":{"value":4295032833}},
+ {"id":197,"raw":{"value":8}},
+ {"id":198,"raw":{"value":8}},
+ {"id":199,"raw":{"value":3}}]}}`
+
+// COMPANION RED-PROOF (observed): delete the three R-330 cases in parseSMART → this fails with
+// "187 Reported_Uncorrect must be carried (raw 1001); got <nil>". Restored.
+func TestParseSMART_R330_CarriesTheThreeCounters(t *testing.T) {
+	s := parseSMART([]byte(r330SATA))
+	if s.ReportedUncorrect == nil || *s.ReportedUncorrect != 1001 {
+		t.Fatalf("187 Reported_Uncorrect must be carried (raw 1001); got %v", s.ReportedUncorrect)
+	}
+	// 188's raw value is vendor-packed on some drives (this one is 0x100010001): carried as reported, not truncated.
+	if s.CommandTimeout == nil || *s.CommandTimeout != 4295032833 {
+		t.Fatalf("188 Command_Timeout must be carried as the full raw value; got %v", s.CommandTimeout)
+	}
+	if s.UDMACRCErrors == nil || *s.UDMACRCErrors != 3 {
+		t.Fatalf("199 UDMA_CRC_Error_Count must be carried (raw 3); got %v", s.UDMACRCErrors)
+	}
+	if s.PendingSectors == nil || *s.PendingSectors != 8 {
+		t.Fatalf("the existing counters must be unchanged; pending=%v", s.PendingSectors)
+	}
+}
+
+// A drive (or an NVMe device) that does not report the attributes leaves them nil, and the JSON
+// OMITS the keys — the receiver reads "unknown", never a measured zero.
+//
+// COMPANION RED-PROOF (observed): drop `,omitempty` from the three tags in hub.SmartSummary → this
+// fails with "an unreported attribute must be omitted, not sent: … reported_uncorrect …". Restored.
+func TestParseSMART_R330_AbsentIsOmittedNotZero(t *testing.T) {
+	for name, raw := range map[string]string{
+		"sata without the three": `{"smart_status":{"passed":true},"ata_smart_attributes":{"table":[{"id":5,"raw":{"value":0}}]}}`,
+		"nvme":                   `{"smart_status":{"passed":true},"nvme_smart_health_information_log":{"critical_warning":0,"media_errors":0,"percentage_used":3}}`,
+	} {
+		s := parseSMART([]byte(raw))
+		if s.ReportedUncorrect != nil || s.CommandTimeout != nil || s.UDMACRCErrors != nil {
+			t.Fatalf("%s: unreported attributes must stay nil; got %v %v %v", name, s.ReportedUncorrect, s.CommandTimeout, s.UDMACRCErrors)
+		}
+		b, _ := json.Marshal(s)
+		for _, k := range []string{"reported_uncorrect", "command_timeout", "udma_crc_errors"} {
+			if strings.Contains(string(b), `"`+k+`"`) {
+				t.Fatalf("%s: an unreported attribute must be omitted, not sent: %s", name, b)
+			}
+		}
+	}
+}

internal/storage/r861_privapply_contract_test.go

@@ -0,0 +1,43 @@
+package storage
+
+import (
+	"strings"
+	"testing"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/privapplytest"
+)
+
+// R-861: every unit the agent renders is one the root checker installs — the name it computes, the Where, the
+// options. RED-PROOF: drop "nosuid","nodev" from mountOptions → the network cases are REFUSED [U5].
+func TestPrivApply_AcceptsTheRenderedUnits(t *testing.T) {
+	local := MountSpec{Name: "x", UUID: "91d2dc2d-2d28-4929-9bdd-3e11fa2f41ae", Where: "/mnt/hdd_1", FSType: "ext4"}
+	name, _ := UnitNameForMount(local.Where)
+	if got := privapplytest.Check(t, "unit", name, renderMountUnit(local)); got != "OK" {
+		t.Errorf("local unit %s: %s", name, got)
+	}
+	local.FSType = ""
+	if got := privapplytest.Check(t, "unit", name, renderMountUnit(local)); got != "OK" {
+		t.Errorf("local unit without Type: %s", got)
+	}
+	for _, spec := range []NetworkMountSpec{
+		{Name: "media", Protocol: ProtocolNFS, Server: "10.0.0.5", Export: "/srv/media", MappedUID: 1000, MappedGID: 1000},
+		{Name: "photos", Protocol: ProtocolSMB, Server: "nas.lan", Export: "photos", CredsRef: "/etc/felhom/netmount/photos.cred", MappedUID: 1000, MappedGID: 1000},
+	} {
+		mu, err := UnitNameForMount(spec.Where())
+		if err != nil {
+			t.Fatal(err)
+		}
+		if got := privapplytest.Check(t, "unit", mu, renderNetworkMountUnit(spec)); got != "OK" {
+			t.Errorf("%s .mount: %s", spec.Name, got)
+		}
+		au := strings.TrimSuffix(mu, ".mount") + ".automount"
+		if got := privapplytest.Check(t, "unit", au, renderNetworkAutomountUnit(spec)); got != "OK" {
+			t.Errorf("%s .automount: %s", spec.Name, got)
+		}
+	}
+	// control: the checker is really looking — a unit over /etc is refused
+	evil := strings.Replace(renderMountUnit(MountSpec{UUID: local.UUID, Where: "/mnt/hdd_1"}), "Where=/mnt/hdd_1", "Where=/etc/sudoers.d", 1)
+	if got := privapplytest.Check(t, "unit", name, evil); !strings.HasPrefix(got, "REFUSED") {
+		t.Fatalf("control: a unit over /etc/sudoers.d was not refused: %s", got)
+	}
+}

internal/storage/smart.go

@@ -44,6 +44,9 @@ const (
 	ataReallocatedSectorCt = 5
 	ataCurrentPending      = 197
 	ataOfflineUncorrect    = 198
+	ataReportedUncorrect   = 187 // R-330
+	ataCommandTimeout      = 188 // R-330
+	ataUDMACRCErrorCount   = 199 // R-330
 )
 
 // parseSMART maps smartctl JSON to a hub.SmartSummary, handling SATA + NVMe and degrading
@@ -85,6 +88,12 @@ func parseSMART(raw []byte) hub.SmartSummary {
 				s.PendingSectors = intPtr(int(a.Raw.Value))
 			case ataOfflineUncorrect:
 				s.OfflineUncorrectable = intPtr(int(a.Raw.Value))
+			case ataReportedUncorrect:
+				s.ReportedUncorrect = int64Ptr(a.Raw.Value)
+			case ataCommandTimeout:
+				s.CommandTimeout = int64Ptr(a.Raw.Value)
+			case ataUDMACRCErrorCount:
+				s.UDMACRCErrors = int64Ptr(a.Raw.Value)
 			}
 		}
 	}
@@ -142,3 +151,5 @@ func parseThinPoolMetadata(raw []byte) (float64, bool) {
 }
 
 func intPtr(v int) *int { return &v }
+
+func int64Ptr(v int64) *int64 { return &v }

internal/wgtunnel/manager.go

@@ -29,6 +29,8 @@ const (
 	Iface = "wg-felhom"
 	// confDest is the installed conf path — must match the FELHOM_WG sudoers entry EXACTLY.
 	confDest = "/etc/wireguard/wg-felhom.conf"
+	// privApply is the root content checker that installs confDest (R-861).
+	privApply = "/usr/local/sbin/felhom-priv-apply"
 	// unit is the systemd unit the sudoers allowlists.
 	unit = "wg-quick@wg-felhom"
 
@@ -507,8 +509,9 @@ func (m *Manager) ensureTunnelLocked(ctx context.Context, block *hub.WireWiregua
 		m.logger.Error("wgtunnel: staging conf", "err", err)
 		return
 	}
-	// argv matches the FELHOM_WG sudoers entry exactly (fixed source + dest).
-	if _, errOut, err := m.runner.Run(ctx, "install", "-o", "root", "-g", "root", "-m", "0600", "--", m.stagedConfPath(), confDest); err != nil {
+	// R-861 (v0.146.0): the root checker installs the staged conf (fixed source /var/lib/felhom-agent/wg/, fixed
+	// destination, 0600) and refuses any key renderConf never writes — PostUp/PreUp run as root under wg-quick.
+	if _, errOut, err := m.runner.Run(ctx, privApply, "wg"); err != nil {
 		m.logger.Error("wgtunnel: conf install failed", "err", err, "stderr", strings.TrimSpace(string(errOut)))
 		return
 	}

internal/wgtunnel/manager_test.go

@@ -286,7 +286,7 @@ func TestScenarioA_RegisterThenApplyOrdering(t *testing.T) {
 	// Block arrives → conf staged + installed + enabled.
 	pub := localPub(t, m)
 	m.Apply(ctx, true, testBlock(pub))
-	if rr.count("install") != 1 || rr.count("systemctl") != 1 {
+	if rr.count(privApply) != 1 || rr.count("systemctl") != 1 {
 		t.Fatalf("apply execs = %v", rr.calls)
 	}
 	if got := rr.lastSystemctl(); strings.Join(got, " ") != "systemctl enable --now wg-quick@wg-felhom" {
@@ -419,7 +419,7 @@ func TestAdoptLostMarkerWithKnownKey(t *testing.T) {
 	if mk == nil || mk.Pubkey != pub || mk.AssignedIP != "10.77.0.2/32" {
 		t.Fatalf("adoption marker = %+v", mk)
 	}
-	if rr.count("install") != 1 {
+	if rr.count(privApply) != 1 {
 		t.Errorf("adopt did not proceed to conf apply: %v", rr.calls)
 	}
 }
@@ -686,13 +686,13 @@ func TestInitialResolveFailureNoTeardown(t *testing.T) {
 	m.Apply(ctx, false, nil) // register (no DNS)
 	pub := localPub(t, m)
 	m.Apply(ctx, true, testBlock(pub)) // resolve fails → cannot apply
-	if rr.count("install") != 0 || rr.count("systemctl") != 0 {
+	if rr.count(privApply) != 0 || rr.count("systemctl") != 0 {
 		t.Errorf("applied/tore-down despite a resolve failure: %v", rr.calls)
 	}
 	// DNS returns → the tunnel comes up on the next tick.
 	fr.set(netip.MustParseAddr("167.233.158.164"))
 	m.Apply(ctx, true, testBlock(pub))
-	if rr.count("install") != 1 {
+	if rr.count(privApply) != 1 {
 		t.Errorf("did not recover after DNS returned: %v", rr.calls)
 	}
 }

internal/wgtunnel/r861_privapply_contract_test.go

@@ -0,0 +1,27 @@
+package wgtunnel
+
+import (
+	"strings"
+	"testing"
+
+	"gitea.dooplex.hu/admin/felhom-agent/internal/privapplytest"
+)
+
+// R-861: the conf renderConf writes (with and without the operator OOB peer) is accepted; a PostUp line is not.
+func TestPrivApply_AcceptsTheRenderedConf(t *testing.T) {
+	priv := "AAECAwQFBgcICQoLDA0ODxAREhMUFRYXGBkaGxwdHh8="
+	b := testBlock("CQkJCQkJCQkJCQkJCQkJCQkJCQkJCQkJCQkJCQkJCQk=")
+	for _, oob := range []string{"", "10.77.0.250"} {
+		b.OOBPeerIP = oob
+		conf, err := renderConf(b, priv, "49.12.1.2")
+		if err != nil {
+			t.Fatal(err)
+		}
+		if got := privapplytest.Check(t, "wg", "", conf); got != "OK" {
+			t.Errorf("rendered conf (oob=%q): %s", oob, got)
+		}
+		if got := privapplytest.Check(t, "wg", "", strings.Replace(conf, "MTU = 1280", "MTU = 1280\nPostUp = id", 1)); !strings.HasPrefix(got, "REFUSED") {
+			t.Fatalf("control: PostUp was not refused: %s", got)
+		}
+	}
+}

scripts/build-step-bundle.py

@@ -0,0 +1,59 @@
+#!/usr/bin/env python3
+"""build-step-bundle.py — the TRANSITION bundle for a release whose bundle ADDS a path (R-880, `11` §5.4.2).
+
+Usage:  python3 scripts/build-step-bundle.py <base-bundle.json> <step-version> <out.json>    (prints the sha256)
+
+WHY. A box's INSTALLED felhom-os-apply checks every path of an incoming bundle against ITS OWN table (rule R16) — so a
+bundle that adds a path (v0.146.1 adds felhom-priv-apply, the guest hook and the shared-parent files, R-861) is refused
+by every box still running an older wrapper. The fix is a step: first a bundle the old wrapper accepts that brings ONLY
+the new felhom-os-apply (the new table), then the release's own bundle, which the new wrapper accepts.
+
+WHAT IT BUILDS. <base-bundle.json> is the bundle the boxes run now (download it from the package registry, e.g.
+felhom-agent/0.145.0/felhom-config-bundle.json, and check its sha against the hub's record). The step bundle is that
+bundle with EXACTLY ONE change: the /usr/local/sbin/felhom-os-apply entry's content is replaced by configs/felhom-os-apply
+(this tree). Same paths, same modes, same checks, every other byte identical; agent_version is <step-version> (e.g.
+0.146.1-step1). The old wrapper verifies it like any bundle (signature, sha, R16, content checks, self-check of the new
+wrapper) — nothing about the trust route changes.
+
+Pinned by configs/test_felhom_config_bundle.py (StepBundle): same paths as the base, only the wrapper differs, the
+new wrapper's table is a superset of the base's paths.
+"""
+import base64
+import hashlib
+import json
+import pathlib
+import re
+import sys
+
+REPO = pathlib.Path(__file__).resolve().parent.parent
+OSAPPLY_DEST = "/usr/local/sbin/felhom-os-apply"
+
+
+def build_step(base_bytes, version, new_osapply_bytes):
+    if not re.match(r"^[0-9]+\.[0-9]+\.[0-9]+-[0-9A-Za-z.]+$", version):
+        raise SystemExit(f"build-step-bundle: {version!r} must be a semver with a step suffix, e.g. 0.146.1-step1")
+    base = json.loads(base_bytes)
+    files = base.get("files")
+    if base.get("format") != 1 or not isinstance(files, list):
+        raise SystemExit("build-step-bundle: the base is not a format-1 bundle")
+    hit = [e for e in files if e.get("path") == OSAPPLY_DEST]
+    if len(hit) != 1:
+        raise SystemExit(f"build-step-bundle: the base has {len(hit)} {OSAPPLY_DEST} entries, want exactly 1")
+    hit[0]["content_b64"] = base64.b64encode(new_osapply_bytes).decode()
+    hit[0]["sha256"] = hashlib.sha256(new_osapply_bytes).hexdigest()
+    base["agent_version"] = version
+    return (json.dumps(base, indent=1, sort_keys=True) + "\n").encode()
+
+
+def main(argv):
+    if len(argv) != 4:
+        print(__doc__, file=sys.stderr)
+        return 2
+    data = build_step(pathlib.Path(argv[1]).read_bytes(), argv[2], (REPO / "configs" / "felhom-os-apply").read_bytes())
+    pathlib.Path(argv[3]).write_bytes(data)
+    print(hashlib.sha256(data).hexdigest())
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main(sys.argv))

scripts/retention-policy.json

@@ -10,13 +10,11 @@
     "CI went red at a commit whose own run had been green the day before, on a true finding that",
     "no one could act on. The red will return at the next publish unless the two read one number.",
     "",
-    "HOW THE NUMBER WAS ARRIVED AT — stated honestly, because it is weaker than it looks.",
-    "generic_versions_kept is 10 because that is what the registry demonstrably holds today",
-    "(felhom-agent 0.121.0..0.128.0 = 10 versions, queried 2026-08-09). It is an OBSERVED state,",
-    "NOT a ruling anyone has been able to locate: no register row records a package prune, R-210",
-    "is WAITING-ON-OPERATOR and says 'Nothing was deleted; this is a list, not an action', and it",
-    "concerns local Docker images rather than this registry. Container packages currently hold 19",
-    "each, so there is no uniform ten-per-package cap visible either. See R-287.",
+    "HOW THE NUMBER WAS ARRIVED AT. generic_versions_kept is 10 because that is what the registry",
+    "keeps: the deleter was ESTABLISHED on 2026-08-10 (R-287) — the newest-10 prune of generic packages",
+    "run under R-267 (felhom-agent and felhom-golden generic held exactly 10 afterwards). Until then this",
+    "file called the 10 an observed state with no located ruling; that is superseded (corrected",
+    "2026-10-05, R-291). Container packages are not pruned by that rule.",
     "",
     "SO THIS FILE IS A FLOOR, NOT A LICENCE. It says: CI may assume nothing older than the newest",
     "N generic versions is still downloadable. It does NOT authorise deleting anything, and the",
@@ -36,9 +34,8 @@
   ],
   "generic_versions_kept": 10,
   "readers": [
-    "scripts/check-published-versions.py — bounds its assertion to the newest N versions",
-    "documentation/runbooks/registry-retention.md (felhom.eu) — the prune procedure"
+    "scripts/check-published-versions.py — bounds its assertion to the newest N versions"
   ],
   "recorded": "2026-08-09",
-  "recorded_by": "CC, from the registry's observed state; NOT from a located operator ruling"
+  "recorded_by": "CC 2026-08-09; the number's source (the R-267 newest-10 prune, established 2026-08-10 by R-287) recorded 2026-10-05 (R-291)"
 }

scripts/test_gate_decoys.py

@@ -0,0 +1,367 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""test_gate_decoys.py — can this repo's gates be fooled by a LABEL? (R-421, R-426)
+
+The same instrument as `felhom.eu/scripts/test_gate_decoys.py`: a decoy is the LABEL without the
+FACT, and a gate that passes on the label alone — or refuses the genuine article — is a live hole.
+Every gate is asserted in BOTH directions: the decoy must be convicted, the genuine article passed.
+
+Covered here (the `COVERS` literal is AST-read by `felhom.eu/scripts/decoy_coverage_gate.py`, which
+never imports this file):
+
+  published         check-published-versions.py, against a FAKE Gitea (see below).
+  release-complete  check-release-complete.py, in a scratch clone whose `origin` is a scratch bare
+                    repository, against the same fake Gitea.
+  reuse-refs, instructions, observations
+                    the three SHARED felhom.eu scripts, run against a scratch clone of THIS repo —
+                    so the decoy is planted in the agent's own REUSE.md / CLAUDE.md / REPORT.md and
+                    coverage is per input, not per script.
+
+NEVER THE REAL GITEA. Both network gates read `GITEA_BASE` from the environment (CI already sets it
+to the in-cluster URL); here it points at an `http.server` bound to 127.0.0.1 inside this process,
+and every proxy variable is removed from the child's environment so urllib cannot route around it.
+A test that asked the real registry would pass or fail on whatever was published that day — the
+constant-for-measurement shape — and would reach the network from a hook.
+
+NEVER THE REAL TREE. Every planted file lives in a scratch directory: a workspace that holds a
+clone of this repo beside symlinks to the sibling clones the shared scripts reach across to.
+
+Run from the repo root:  python3 scripts/test_gate_decoys.py
+Exit 0 every decoy judged correctly · 1 a decoy passed or a genuine article was refused.
+"""
+import http.server
+import io
+import json
+import os
+import shutil
+import socketserver
+import subprocess
+import sys
+import tempfile
+import threading
+
+ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+PARENT = os.path.dirname(ROOT)
+# DECOY_SHARED_DIR exists for ONE purpose: the red-proof. It lets a mutated COPY of the shared scripts
+# be judged without editing the felhom.eu clone. Unset, the suite judges the real shared scripts.
+SHARED = os.environ.get("DECOY_SHARED_DIR") or os.path.join(PARENT, "felhom.eu", "scripts")
+
+# ── WHAT THIS FILE COVERS ────────────────────────────────────────────────────────────────────────
+# Read by felhom.eu/scripts/decoy_coverage_gate.py, which AST-parses this literal. A gate named here
+# MUST have a decoy below that has been seen to fail.
+COVERS = {
+    "published": "against a FAKE Gitea: a tag whose package 404s, a tag whose tree lacks the configs, a package one "
+                 "version past the newest tag (published, never tagged), a patch-gap orphan, a missing package that "
+                 "lexical sorting would drop out of the retention window (0.9.x vs 0.10.x); a tags api answering 500 "
+                 "or a non-JSON 200 is INCONCLUSIVE, never a pass - vs a clean registry, a non-semver tag and a version "
+                 "older than the retention window (not asserted, BY DESIGN) (R-426)",
+    "release-complete": "the newest `## vX.Y.Z` with no tag anywhere, a tag parked on an unrelated commit, a tag "
+                        "with no package; a registry 500 is INCONCLUSIVE - vs the genuine release, an `## Unreleased` "
+                        "heading above it, a newer version named only in prose or under `###`, a tag that only "
+                        "origin has (the shallow-CI shape); a LOCAL-only tag passes BY DESIGN (CI's fresh clone "
+                        "and the published gate's converse probe are what see it) (R-426)",
+    "reuse-refs": "a cited .go and a cited .md path that do not exist, planted in THIS repo's REUSE.md - vs the "
+                  "real file (R-426)",
+    "instructions": "a component version literal in THIS repo's CLAUDE.md effective text - vs the same sentence "
+                    "inside an HTML comment (R-426)",
+    "observations": "R-419 in THIS repo's REPORT.md: an Observations note SAYING it carries no marker - vs the "
+                    "two genuine markers (R-426)",
+}
+
+fails = []
+ran = 0
+
+
+def report(name, rc, out, expect_rc, must=()):
+    global ran
+    ran += 1
+    missing = [m for m in must if m not in out]
+    if rc == expect_rc and not missing:
+        print("  ok  %-62s rc=%d (expected %d)" % (name, rc, expect_rc))
+    else:
+        hole = expect_rc != 0 and rc == 0
+        fails.append("%s: rc=%d expected %d%s; missing %s\n%s" % (
+            name, rc, expect_rc, " - LIVE HOLE" if hole else "", missing, out[-900:]))
+
+
+# ── the fake Gitea ───────────────────────────────────────────────────────────────────────────────
+class Fake(object):
+    """What the fake registry serves. Reset per case."""
+
+    def reset(self):
+        self.tags = []            # tag names, as the tags api lists them
+        self.packages = set()     # versions whose generic package downloads
+        self.raw = set()          # versions whose tag tree serves the probe config
+        self.tags_status = 200
+        self.tags_body = None     # override bytes for the tags api
+        self.pkg_status = None    # override status for EVERY package request
+        self.hits = []
+
+
+FAKE = Fake()
+FAKE.reset()
+PKG_PREFIX = "/api/packages/admin/generic/felhom-agent/"
+RAW_PREFIX = "/admin/felhom-agent/raw/tag/v"
+
+
+class Handler(http.server.BaseHTTPRequestHandler):
+    def log_message(self, *a):
+        pass
+
+    def _answer(self, status, body=b""):
+        self.send_response(status)
+        self.send_header("Content-Length", str(len(body)))
+        self.end_headers()
+        if self.command != "HEAD":
+            self.wfile.write(body)
+
+    def do_GET(self):
+        p = self.path
+        FAKE.hits.append(p)
+        if p.startswith("/api/v1/repos/admin/felhom-agent/tags"):
+            body = FAKE.tags_body if FAKE.tags_body is not None else \
+                json.dumps([{"name": t} for t in FAKE.tags]).encode()
+            return self._answer(FAKE.tags_status, body)
+        if p.startswith(PKG_PREFIX):
+            if FAKE.pkg_status is not None:
+                return self._answer(FAKE.pkg_status)
+            v = p[len(PKG_PREFIX):].split("/", 1)[0]
+            return self._answer(200, b"ELF") if v in FAKE.packages else self._answer(404)
+        if p.startswith(RAW_PREFIX):
+            v = p[len(RAW_PREFIX):].split("/", 1)[0]
+            ok = v in FAKE.raw and p.endswith("/configs/felhom-agent.service")
+            return self._answer(200, b"[Unit]\n") if ok else self._answer(404)
+        return self._answer(404)
+
+    do_HEAD = do_GET
+
+
+class Server(socketserver.ThreadingMixIn, http.server.HTTPServer):
+    daemon_threads = True
+
+
+def child_env(base):
+    env = {k: v for k, v in os.environ.items() if "proxy" not in k.lower()}
+    env["GITEA_BASE"] = base
+    env["NO_PROXY"] = env["no_proxy"] = "127.0.0.1,localhost"
+    return env
+
+
+def run(argv, cwd, env=None):
+    # input="" — a child must never inherit (and block on) this process's stdin
+    p = subprocess.run(argv, cwd=cwd, env=env, capture_output=True, text=True, input="")
+    return p.returncode, p.stdout + p.stderr
+
+
+def sh(argv, cwd):
+    rc, out = run(argv, cwd)
+    if rc != 0:
+        raise SystemExit("setup command failed (%s): %s" % (" ".join(argv), out))
+    return out.strip()
+
+
+# ── published ────────────────────────────────────────────────────────────────────────────────────
+def published_cases(base):
+    gate = os.path.join(ROOT, "scripts", "check-published-versions.py")
+    keep = json.load(io.open(os.path.join(ROOT, "scripts", "retention-policy.json"),
+                             encoding="utf-8"))["generic_versions_kept"]
+    TAGS = ["v0.150.%d" % i for i in range(3)]          # inside any retention window >= 3
+    VERS = [t[1:] for t in TAGS]
+
+    def case(name, setup, expect_rc, must=()):
+        FAKE.reset()
+        FAKE.tags = list(TAGS)
+        FAKE.packages = set(VERS)
+        FAKE.raw = set(VERS)
+        setup()
+        rc, out = run([sys.executable, gate], ROOT, child_env(base))
+        if not FAKE.hits:
+            fails.append("published/%s: the gate never asked the fake Gitea - the seam is not wired" % name)
+        report("published: " + name, rc, out, expect_rc, must)
+
+    case("GENUINE: every tag downloadable and serving its configs", lambda: None, 0,
+         ("ALL RELEASED VERSIONS INSTALLABLE",))
+    case("GENUINE: a non-semver tag is not a release", lambda: FAKE.tags.append("v0.150.2-rc1"), 0,
+         ("ALL RELEASED VERSIONS INSTALLABLE",))
+    case("FACT: a tag whose package 404s", lambda: FAKE.packages.discard("0.150.1"), 1,
+         ("FAIL v0.150.1", "binary NOT downloadable"))
+    case("FACT: a tag whose tree does not serve the configs", lambda: FAKE.raw.discard("0.150.2"), 1,
+         ("FAIL v0.150.2", "does not serve"))
+    case("FACT: published one patch past the newest tag, never tagged",
+         lambda: FAKE.packages.add("0.150.3"), 1, ("PUBLISHED VERSION(S) WITH NO TAG", "v0.150.3"))
+    case("FACT: published in a patch GAP between two tags",
+         lambda: (FAKE.tags.remove("v0.150.1"),), 1, ("v0.150.1 is downloadable", "has no git tag"))
+
+    def lexical():
+        # keep+1 tags: 0.9.0 and 0.10.0..0.10.<keep-1>. By SEMVER the oldest is 0.9.0 (dropped); by
+        # STRING sort "0.10.0" is the smallest and would be the one dropped - so its missing package
+        # is convicted only if the window is cut by semver.
+        FAKE.tags = ["v0.9.0"] + ["v0.10.%d" % i for i in range(keep)]
+        FAKE.packages = set(t[1:] for t in FAKE.tags) - {"0.10.0"}
+        FAKE.raw = set(t[1:] for t in FAKE.tags)
+    case("FACT: a missing package lexical sorting would drop (0.10.0 vs 0.9.0)", lexical, 1,
+         ("FAIL v0.10.0",))
+
+    def retired():
+        FAKE.tags = ["v0.9.0"] + ["v0.10.%d" % i for i in range(keep)]
+        FAKE.packages = set(t[1:] for t in FAKE.tags) - {"0.9.0"}
+        FAKE.raw = set(t[1:] for t in FAKE.tags)
+    case("BY DESIGN: a version older than the retention window is not asserted", retired, 0,
+         ("NOT ASSERTED", "0.9.0"))
+
+    def five_hundred():
+        FAKE.tags_status = 500
+    case("INCONCLUSIVE: the tags api answers 500", five_hundred, 2, ("INCONCLUSIVE",))
+
+    def html():
+        FAKE.tags_body = b"<html>sign in</html>"
+    case("INCONCLUSIVE: the tags api answers a 200 that is not JSON", html, 2, ("INCONCLUSIVE",))
+
+    # an unreachable Gitea: a port nothing listens on
+    s = Server(("127.0.0.1", 0), Handler)
+    dead = "http://127.0.0.1:%d" % s.server_address[1]
+    s.server_close()
+    rc, out = run([sys.executable, gate], ROOT, child_env(dead))
+    report("published: INCONCLUSIVE: Gitea unreachable", rc, out, 2, ("INCONCLUSIVE", "URLs tried"))
+
+
+# ── release-complete ─────────────────────────────────────────────────────────────────────────────
+def release_cases(base, ws):
+    bare = os.path.join(ws, "origin.git")
+    work = os.path.join(ws, "rc-work")
+    sh(["git", "clone", "-q", "--bare", "--no-tags", "file://" + ROOT, bare], ws)
+    sh(["git", "clone", "-q", "--no-tags", "file://" + bare, work], ws)
+    sh(["git", "config", "user.email", "decoy@gate.invalid"], work)
+    sh(["git", "config", "user.name", "decoy"], work)
+    # the WORKING-TREE gate, so the file under test is the one being edited, not HEAD's
+    shutil.copy(os.path.join(ROOT, "scripts", "check-release-complete.py"),
+                os.path.join(work, "scripts", "check-release-complete.py"))
+    sh(["git", "add", "scripts/check-release-complete.py"], work)
+    sh(["git", "commit", "-q", "--allow-empty", "-m", "the gate under test"], work)
+    base_sha = sh(["git", "rev-parse", "HEAD"], work)
+    ch = os.path.join(work, "CHANGELOG.md")
+    original = io.open(ch, encoding="utf-8").read()
+    V = "9.9.9"
+
+    def case(name, top, expect_rc, must=(), tag=None, origin_tag=False, packaged=True, pkg_status=None):
+        FAKE.reset()
+        if packaged:
+            FAKE.packages = {V}
+        FAKE.pkg_status = pkg_status
+        try:
+            io.open(ch, "w", encoding="utf-8").write(top + original)
+            sh(["git", "commit", "-q", "-am", name], work)
+            if tag == "head":
+                sh(["git", "tag", "-a", "v" + V, "-m", "decoy", "HEAD"], work)
+            elif tag == "unrelated":
+                empty = sh(["git", "mktree"], work)   # stdin is "" — the empty tree, written to this repo
+                orphan = sh(["git", "commit-tree", "-m", "unrelated", empty], work)
+                sh(["git", "tag", "-a", "v" + V, "-m", "decoy", orphan], work)
+            if origin_tag:
+                sh(["git", "push", "-q", "origin", "HEAD:refs/tags/v" + V], work)
+            rc, out = run([sys.executable, os.path.join(work, "scripts", "check-release-complete.py")],
+                          work, child_env(base))
+            report("release-complete: " + name, rc, out, expect_rc, must)
+        finally:
+            run(["git", "tag", "-d", "v" + V], work)
+            run(["git", "push", "-q", "origin", ":refs/tags/v" + V], work)
+            sh(["git", "reset", "-q", "--hard", base_sha], work)
+
+    HEAD = "## v%s — 2026-10-06\n\n- decoy release\n\n" % V
+    case("GENUINE: tagged at HEAD and published", HEAD, 0,
+         ("newest CHANGELOG version: v9.9.9", "is tagged, placed and published"), tag="head")
+    case("GENUINE: an `## Unreleased` heading above the release", "## Unreleased\n\n- wip\n\n" + HEAD, 0,
+         ("newest CHANGELOG version: v9.9.9",), tag="head")
+    case("GENUINE: a newer version named only in prose and under ###",
+         "The `## v10.0.0` heading is not written yet.\n### v10.0.0 notes\n\n" + HEAD, 0,
+         ("newest CHANGELOG version: v9.9.9",), tag="head")
+    case("GENUINE: the tag only on origin (the shallow-CI shape)", HEAD, 0,
+         ("exists on origin",), origin_tag=True)
+    case("BY DESIGN: a LOCAL-only tag passes (CI's fresh clone sees only origin)", HEAD, 0,
+         ("an ancestor of HEAD",), tag="head")
+    case("FACT: the newest heading has no tag anywhere", HEAD, 1, ("DOES NOT EXIST",))
+    case("FACT: a tag parked on an unrelated commit", HEAD, 1, ("NOT an ancestor",), tag="unrelated")
+    case("FACT: tagged, never published", HEAD, 1, ("IS NOT PUBLISHED",), tag="head", packaged=False)
+    case("INCONCLUSIVE: the registry answers 500", HEAD, 2, ("INCONCLUSIVE",), tag="head", pkg_status=500)
+    case("FACT beats INCONCLUSIVE: no tag AND the registry answers 500", HEAD, 1, ("DOES NOT EXIST",),
+         pkg_status=500)
+
+
+# ── the shared felhom.eu scripts, against THIS repo's inputs ─────────────────────────────────────
+def shared_cases(ws):
+    """A scratch WORKSPACE: a clone of this repo beside symlinks to the siblings, because the shared
+    scripts reach across (REUSE.md cites hub paths; instructions_gate reads the workspace CLAUDE.md)."""
+    for g in ("reuse_refs_check.py", "instructions_gate.py", "observations_gate.py"):
+        if not os.path.isfile(os.path.join(SHARED, g)):
+            fails.append("shared gate %s is MISSING beside this clone (tried %s) - a failure, never a skip"
+                         % (g, SHARED))
+            return
+    space = os.path.join(ws, "workspace")
+    os.makedirs(space)
+    for entry in sorted(os.listdir(PARENT)):
+        if entry in ("felhom.eu", "felhom-controller", "app-catalog-felhom.eu", "homelab-manifests",
+                     "CLAUDE.md", ".claude-memory"):
+            os.symlink(os.path.join(PARENT, entry), os.path.join(space, entry))
+    repo = os.path.join(space, "felhom-agent")
+    sh(["git", "clone", "-q", "--no-tags", "file://" + ROOT, repo], ws)
+    # the WORKING-TREE inputs the plants go into, so a case judges today's file
+    for f in ("REUSE.md", "CLAUDE.md", "REPORT.md"):
+        shutil.copy(os.path.join(ROOT, f), os.path.join(repo, f))
+
+    def case(name, gate, relpath, extra, expect_rc, must=()):
+        p = os.path.join(repo, relpath)
+        backup = io.open(p, encoding="utf-8").read()
+        try:
+            if extra:
+                io.open(p, "w", encoding="utf-8").write(backup + extra)
+            rc, out = run([sys.executable, os.path.join(SHARED, gate), repo], repo)
+            report(name, rc, out, expect_rc, must)
+        finally:
+            io.open(p, "w", encoding="utf-8").write(backup)
+
+    case("reuse-refs: GENUINE: this repo's REUSE.md", "reuse_refs_check.py", "REUSE.md", "", 0, ("FAILED 0",))
+    case("reuse-refs: FACT: a cited .go path that does not exist", "reuse_refs_check.py", "REUSE.md",
+         u"\n- see `internal/localapi/does_not_exist.go`\n", 1, ("does_not_exist.go",))
+    case("reuse-refs: FACT: a cited .md path that does not exist", "reuse_refs_check.py", "REUSE.md",
+         u"\n- see `docs/99-does-not-exist.md`\n", 1, ("99-does-not-exist.md",))
+    case("instructions: GENUINE: this repo's CLAUDE.md", "instructions_gate.py", "CLAUDE.md", "", 0,
+         ("instructions_gate: OK",))
+    case("instructions: FACT: a version literal in effective text", "instructions_gate.py", "CLAUDE.md",
+         u"\nThe agent runs v0.148.0 today.\n", 1, ("v0.148.0",))
+    case("instructions: GENUINE: the same sentence in an HTML comment", "instructions_gate.py", "CLAUDE.md",
+         u"\n<!--\nThe agent ran v0.148.0 on 2026-10-06.\n-->\n", 0, ("instructions_gate: OK",))
+    case("observations: FACT: R-419, prose SAYING it has no marker", "observations_gate.py", "REPORT.md",
+         u"\n## Observations\n\n1. **A real finding.** It carries no `FILED:` marker and no "
+         u"`NOT-A-FINDING:` marker, deliberately.\n", 1)
+    case("observations: GENUINE: a FILED marker", "observations_gate.py", "REPORT.md",
+         u"\n## Observations\n\n1. **A real finding.** Something broke. **FILED: R-419**\n", 0)
+    case("observations: GENUINE: a NOT-A-FINDING marker", "observations_gate.py", "REPORT.md",
+         u"\n## Observations\n\n1. **A real finding.** Odd. **NOT-A-FINDING: my own typo, corrected in "
+         u"the same minute.**\n", 0)
+
+
+def main():
+    srv = Server(("127.0.0.1", 0), Handler)
+    threading.Thread(target=srv.serve_forever, daemon=True).start()
+    base = "http://127.0.0.1:%d" % srv.server_address[1]
+    ws = tempfile.mkdtemp(prefix="agent-decoys-")
+    print("agent gate decoys — fake Gitea at %s, scratch %s" % (base, ws))
+    try:
+        published_cases(base)
+        release_cases(base, ws)
+        shared_cases(ws)
+    finally:
+        srv.shutdown()
+        srv.server_close()
+        shutil.rmtree(ws, ignore_errors=True)
+    if fails:
+        print()
+        for f in fails:
+            print("FAIL: %s" % f)
+        return 1
+    print("\nagent gate decoys OK — %d case(s), every label judged on its fact (R-421)" % ran)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())