v0.201.0 — a correct recovery code is never called wrong again (CAMPAIGN-11) — MinAgent 0.125.0

R-216: the offsite key recovery is a coupled feature and now says so. featureProbes +
featureMinAgent 0.125.0 + a Supports gate at the unlock entry point, FAILING CLOSED — an
agent that cannot answer is named as such instead of the customer's code being blamed.
Measured live: a 404 from agent 0.120.0 came back as "we did not accept your recovery
code, check that all ten words", in 0.134 s, against a perfect code.

R-218: delete the repo-password short-circuit in needsOffsiteCredential. The declaration
stops when the TIER WORKS, not when a key exists — installing a key is the recovery
screen's whole job, so succeeding at recovery was switching off the mechanism that would
have delivered the coordinates to use it.

R-219: the unlock finishes the job — place the key, bring the tier up, then list. Without
it the promised listing could never render on the shape the screen exists for.

R-217: an unreadable store no longer claims to have opened with unattributable content
(the OffsiteInventory{} zero value). Opened / empty / unreadable are three states.

R-222: a code that is right about a RETAINED earlier package is named, not blamed. States
what the hub knows and promises nothing — no read path exists.

R-215: GET /recovery is gated on the same predicate as the interception.

Five red-proofs, each demonstrated failing and restored.

CHANGELOG.md

@@ -1,3 +1,63 @@
+## v0.201.0 — a correct recovery code is never called wrong again (2026-08-05, CAMPAIGN-11) — MinAgent 0.125.0
+
+CAMPAIGN-11 walked the whole recovery journey end to end for the first time. **The data came back
+byte-identical; the journey did not exist** — four operator interventions stood between a customer and
+their files, and the first one was the machine telling them their perfectly correct recovery code was
+wrong. This closes six of the nine findings.
+
+**R-216 — the headline. A 404 was reported to the customer as a bad code.** The unseal lives in the
+agent behind `POST /escrow/recover-offsite-password`, which ships in agent **v0.125.0**. On an older
+agent the route 404s, and the unlock was attempted anyway: *„A megadott helyreállítási kódot nem
+fogadtuk el. Ellenőrizd, hogy mind a tíz szót pontosan…"* — measured live in 0.134 s, far too fast for
+`age`'s scrypt, against a code that was perfect. **It was the default state, not a misconfiguration**:
+the day-0 manifest vouches agent 0.120.0, and a reinstall actively *downgrades* a hand-fixed box back
+to it. Three parts:
+
+- `FeatureOffsiteKeyRecovery` now exists in `featureProbes` + `featureMinAgent` (0.125.0) with a
+  `Supports` gate at the unlock entry point — **the convention `features.go:72-76` and
+  `publish-train-rules.md` rule 4 already required, which v0.195.0/v0.196.0 declared in their headers
+  and v0.200.0 dropped when it put the same coupled feature behind a customer-facing button.**
+- **This gate FAILS CLOSED**, alone in that table: anything but `SupportYes` says *the machine* cannot
+  ask yet, and no unlock is attempted. An attempt that cannot succeed must never be made, because its
+  failure is what got attributed to the code. The asymmetry is documented beside the row.
+- The hub's own guard could not catch it (see the hub changelog, v0.97.0).
+
+**R-218 — succeeding at recovery stopped the box asking for what it still needed.**
+`needsOffsiteCredential` short-circuited on a repository password existing — and installing one is the
+recovery screen's whole job. 32 seconds after the hub re-staged the credential, the customer's success
+switched off the mechanism that would have delivered the coordinates for the key they had just
+recovered. **The short-circuit is deleted**; the declaration now stops when the tier works, not when a
+key exists. Scenario E (a deliberately disabled target stays silent) is pinned by its own test.
+
+**R-219 — the promised listing could never render on the shape the screen exists for.** Listing needs
+a target; a target cannot exist without a repository password; shape (a) is defined by having none. The
+unlock now **finishes the job** — place the key, bring the tier up, then list — and a tier that is not
+up yet says *waiting for the connection details* rather than reporting a failure.
+
+**R-217 — an unreadable store claimed to have opened.** The failure path passed
+`backup.OffsiteInventory{}`, whose `Empty=false` the template read as *„A tároló megnyílt, és van benne
+tartalom, de nem tudtuk alkalmazásokhoz rendelni."* — three assertions, none of them known. The field
+built to prevent exactly this says so in its own doc comment. Opened / empty / unreadable are now three
+distinguishable states.
+
+**R-222 — reaching for a retained earlier package read as a wrong code.** The hub's ACK carries
+`superseded_present` / `superseded_at` (hub v0.97.0); the screen now names that situation. It states
+the two facts the hub knows and **does not promise the earlier package can be opened** — that read path
+does not exist.
+
+**R-215 — `GET /recovery` rendered the recovery story on a box that never had backups.** The predicate
+was right and the page never asked it. Gated on the same predicate as the interception.
+
+**Four messages where there was one**, each saying what to do next and only one mentioning typing.
+
+Left open deliberately: **R-214** (the console never stops showing a stale pairing code), **R-220**
+(drives cannot be re-enrolled after a rebuild), **R-221** (a rebuilt box cannot run the escrow ceremony
+— a real blocker for that flow).
+
+Tests: `recovery_gate_test.go` (Scenarios A/F/G/H/I, handler-level), `offbox_declare_test.go`
+(Scenarios D/E). **Five red-proofs, each demonstrated failing and restored** — removing the capability
+gate brings the accusation back verbatim.
+
 ## v0.200.0 — the recovery screen: unlocking, and only unlocking (2026-08-05, R-193)
 
 R-193's remaining half. The drill proved a customer's file comes back after a machine is destroyed;

controller/cmd/controller/main.go

@@ -494,6 +494,10 @@ func main() {
 	// updater's floor/update-running state and defer the one-time-password consume past a managed
 	// day-0 floor-update (the F10 race). The gate is wired ONLY when an updater exists — with no update
 	// mechanism there is no floor-update to race, so the bridge reconciles immediately (Settle nil).
+	// offsiteBridge is hoisted out of the block below ONLY so the recovery screen can drive one
+	// reconcile between placing a recovered key and reading the repository (R-219). nil when off-site
+	// is not configured for this customer, which the web seam treats as "skip".
+	var offsiteBridge *offsiteapply.Bridge
 	if backupMgr != nil && cfg.Offsite.Enabled && cfg.Hub.URL != "" && cfg.Hub.APIKey != "" {
 		bridge := &offsiteapply.Bridge{
 			Cfg:       cfg,
@@ -526,6 +530,7 @@ func main() {
 				return Version, floor, u.IsUpdateRunning(), floor != ""
 			})
 		}
+		offsiteBridge = bridge
 		go func() {
 			// ReconcileWhenSettled runs the settle-gate FIRST (its own bounds), then Reconcile under a
 			// fresh 3-minute context — the gate's wait never eats the reconcile budget.
@@ -670,6 +675,9 @@ func main() {
 			// is precisely the case that needs it. Half of the stranded-rebuild predicate in
 			// backup.needsOffsiteCredential; the other half (no repository password) is local.
 			RecordPresence: sett.SetHubEscrowIdentityPresent,
+			// v0.201.0 (R-222): whether the hub is ALSO keeping an EARLIER sealed package, so the
+			// recovery screen can name that situation instead of blaming the customer's typing.
+			RecordSuperseded: sett.SetHubEscrowSuperseded,
 			Logger:           logger,
 		}
 		// Wire hub verification: update settings when hub reports customer status
@@ -1268,6 +1276,13 @@ func main() {
 		// before a code is entered. A timestamp, never a secret.
 		webServer.SetEscrowSealedAt(escrowConfirmer.SealedAt)
 	}
+	// R-219 / R-218: the unlock brings the off-site tier up between placing the recovered key and
+	// reading the repository. Without it the listing the screen promises can never render on the
+	// pristine rebuilt box — no key ⇒ no target ⇒ no inventory — and the apply-bridge's only other
+	// trigger is a config refresh or a restart, neither of which the customer can cause.
+	if offsiteBridge != nil {
+		webServer.SetRecoveryTierUp(offsiteBridge.Reconcile)
+	}
 	webServer.SetIntegrationManager(integrationMgr)
 	if reportTrigger != nil {
 		// Out-of-cycle report push after hub-relevant user actions (escrow claim, settings

controller/internal/agentapi/features.go

@@ -35,6 +35,26 @@ const FeatureGuestMemoryResize Feature = "guest_memory_resize"
 // an empty field means legacy.
 const FeatureBackupAgeState Feature = "backup_age_state"
 
+// FeatureOffsiteKeyRecovery is the customer-facing off-site key recovery (agent v0.125.0, R-199
+// links 7–8): POST /escrow/recover-offsite-password fetches this host's sealed bundle, unseals it
+// with R and returns the single repository-password field.
+//
+// ⚠ THIS GATE FAILS CLOSED, and it is the ONLY feature in this table that does. Read §7.1 of the
+// R-216 fix before "correcting" it back to the package default.
+//
+// The package default is fail-OPEN: SupportUnknown proceeds, because for every other coupled feature
+// a wrong "unsupported" would block something harmless while a down agent already speaks through the
+// normal error paths. **That default is what produced R-216.** Measured live on 2026-08-05
+// (CAMPAIGN-11 Phase 1): an agent 0.120.0 answered the recovery route with 404, the unlock attempt
+// went ahead anyway, and the customer was told — in Hungarian, on the one screen whose whole purpose
+// is to be believed about backups — that their perfectly correct recovery code was not accepted and
+// they should check their typing. A correct code, refused in 0.134 s, blamed on the customer.
+//
+// So here: anything other than SupportYes means the screen says THE MACHINE cannot ask yet. The
+// unlock is never attempted when it cannot complete, because the failure of an attempt that could
+// never have worked is attributed to the code.
+const FeatureOffsiteKeyRecovery Feature = "offsite_key_recovery"
+
 // SupportState is a probe verdict. The zero value is SupportUnknown (fail-open: unknown never
 // refuses — the existing agent-error paths speak honestly when the agent is down).
 type SupportState int
@@ -92,12 +112,27 @@ var featureProbes = map[Feature]func(ctx context.Context, p SupportProber) error
 		_, err := gm.GuestMemory(ctx)
 		return err
 	},
+	// The recovery route is a POST that performs work and consumes a recovery code — it cannot be
+	// probed. Like the memory prober's negative case this returns a sentinel that classifies to
+	// SupportUnknown, so the decision falls to the VERSION path above.
+	//
+	// The row must exist even though it cannot probe: SupportsWithSource looks up featureProbes
+	// FIRST and returns "unregistered"/SupportUnknown on a table gap, before the version path runs.
+	// A featureMinAgent row without a featureProbes row is therefore never consulted at all.
+	FeatureOffsiteKeyRecovery: func(ctx context.Context, p SupportProber) error {
+		return errNoRecoveryProbe
+	},
 }
 
 // errNoMemoryProbe classifies to SupportUnknown (not a *StatusError 404), so a prober that cannot be
 // asked never reads as "unsupported".
 var errNoMemoryProbe = errors.New("agentapi: prober does not support the guest-memory probe")
 
+// errNoRecoveryProbe classifies to SupportUnknown: the off-site key recovery route is a POST that
+// consumes a recovery code and so cannot be probed, leaving the VERSION path to decide. Its caller
+// fails CLOSED on Unknown — see FeatureOffsiteKeyRecovery.
+var errNoRecoveryProbe = errors.New("agentapi: the offsite key recovery route cannot be probed")
+
 // featureMinAgent maps each coupled feature to the MINIMUM agent version that carries its coupled
 // semantics (the CHANGELOG `MinAgent:` header value). Used by Supports when the agent's version is
 // KNOWN (the v0.82.0 X-Felhom-Agent-Version channel) — a direct comparison, no probe traffic. A
@@ -107,8 +142,16 @@ var featureMinAgent = map[Feature]string{
 	FeatureGuestMemoryResize: "0.90.0",
 	// R-88 Part 2: /backup/due carries age_state, distinguishing "never backed up" from "cannot tell".
 	FeatureBackupAgeState: "0.105.0",
+	// R-199 links 7–8: POST /escrow/recover-offsite-password. R-216 — this row is the whole reason a
+	// correct recovery code can no longer be reported as wrong on an agent that cannot answer.
+	FeatureOffsiteKeyRecovery: "0.125.0",
 }
 
+// MinAgentFor returns the declared minimum agent version for a feature ("" when the feature has no
+// row). Read-only accessor over featureMinAgent so a refusal can NAME the version it needs instead of
+// hard-coding the number a second time at the call site.
+func MinAgentFor(f Feature) string { return featureMinAgent[f] }
+
 // AgentVersionReporter is optionally implemented by a SupportProber (*Client is one): it reports
 // the last strictly-validated agent version seen on its traffic ("" = unknown → probe fallback).
 type AgentVersionReporter interface {

controller/internal/backup/offbox.go

@@ -1228,6 +1228,32 @@ const OffsiteStateNeedsCredential = "needs_credential"
 // it". A target that exists but is DISABLED is not stranded either — the customer switched it off —
 // so the caller only consults this when there is no enabled target, and a non-nil disabled target
 // short-circuits to false here.
+//
+// ⚠ THE DECLARATION STOPS WHEN THE TIER WORKS, NOT WHEN A KEY EXISTS (R-218, v0.201.0).
+//
+// This used to carry a third condition: hold a repository password ⇒ not stranded, stop declaring.
+// It reads as a sound freshness test and it is the exact opposite on the one path that matters,
+// because installing a repository password is the RECOVERY SCREEN'S WHOLE JOB. Measured live on
+// 2026-08-05 (CAMPAIGN-11 Phase 1):
+//
+//	13:39:54  needs_credential          the box asks
+//	13:42:43  needs_credential          the hub's 2-report debounce is satisfied
+//	13:47:03  hub re-stages the credential — "the box re-consumes on its next cycle"
+//	13:47:35  the customer's unlock succeeds and places the recovered key
+//	13:53:42  (silence)                 …and never asks again
+//
+// Thirty-two seconds after the remedy fired, the customer's own success switched off the mechanism
+// that would have delivered the coordinates for the key they had just recovered. The hub held an
+// unconsumed credential the box had no reason to collect; the box held a correct key and nowhere to
+// use it; the screen said "a few minutes"; nothing was ever going to happen. Two features, each
+// correct alone, cancelling on the path both were built for.
+//
+// The two remaining conditions are the honest ones: no target at all, and the hub is keeping a sealed
+// package for us. Both stay true exactly until the tier is configured — which is the moment the box
+// genuinely no longer needs a credential — and `OffboxReportStatus` stops consulting this predicate
+// the instant a target exists. **Scenario E is unaffected and is pinned by its own test**: a disabled
+// target is non-nil and still short-circuits at the first line, so a box whose customer switched
+// off-site off stays silent.
 func (m *Manager) needsOffsiteCredential(t *settings.OffboxTarget) bool {
 	if t != nil {
 		return false // a target exists (merely disabled) — the customer's own choice, not a rebuild
@@ -1235,9 +1261,6 @@ func (m *Manager) needsOffsiteCredential(t *settings.OffboxTarget) bool {
 	if m.settings == nil || !m.settings.GetHubEscrowIdentityPresent() {
 		return false // the hub holds nothing for us: never had off-site backups
 	}
-	if _, ok := m.OffboxRepoPasswordHash(); ok {
-		return false // we still hold our repository password: not a fresh data area
-	}
 	return true
 }
 

controller/internal/backup/offbox_declare_test.go

@@ -88,23 +88,60 @@ func TestOffsiteDeclare_NeverHadOffsiteSaysNothing(t *testing.T) {
 	}
 }
 
-// The other half of the conjunction: a box that still holds its repository password is NOT stranded,
-// even though the hub holds an escrow for it. That is simply a healthy box between configurations.
-func TestOffsiteDeclare_BoxThatStillHoldsItsRepoPasswordDoesNotDeclare(t *testing.T) {
+// SCENARIO D (R-218) — THE DECLARATION STOPS WHEN THE TIER WORKS, NOT WHEN A KEY EXISTS.
+//
+// ⚠ THIS TEST ASSERTED THE OPPOSITE until v0.201.0, and it was green the whole time. It required a
+// box holding a repository password to stay SILENT — which reads as a sound freshness test and is the
+// exact opposite on the one path that matters, because installing a repository password is the
+// RECOVERY SCREEN'S WHOLE JOB. Measured live 2026-08-05 (CAMPAIGN-11 Phase 1): 32 seconds after the
+// hub re-staged the credential, the customer's successful unlock switched off the mechanism that
+// would have delivered the coordinates for the key they had just recovered. Deadlock, both halves.
+//
+// RED-PROOF: restore the `if _, ok := m.OffboxRepoPasswordHash(); ok { return false }` short-circuit
+// in needsOffsiteCredential and this test FAILS — the box goes silent again with no target, which is
+// the deadlock. Demonstrated failing before this test was kept.
+func TestOffsiteDeclare_StillDeclaresAfterARecoveredKeyIsPlaced(t *testing.T) {
 	m, sett := bareManager(t)
 	if err := sett.SetHubEscrowIdentityPresent(true); err != nil {
 		t.Fatal(err)
 	}
-	// Place a repository password exactly where the manager looks for it.
+	// The post-unlock shape: the recovered repository password is on disk, and there is STILL no
+	// off-site target — so the box cannot use what it just recovered.
 	if err := os.MkdirAll(m.offboxDir(), 0o700); err != nil {
 		t.Fatal(err)
 	}
-	if err := os.WriteFile(m.offboxPwPath(), []byte("a-repository-password"), 0o600); err != nil {
+	if err := os.WriteFile(m.offboxPwPath(), []byte("a-recovered-repository-password"), 0o600); err != nil {
 		t.Fatal(err)
 	}
 
-	if st := m.OffboxReportStatus(); st != nil {
-		t.Fatalf("a box holding its repository password declared a need: %+v", st)
+	st := m.OffboxReportStatus()
+	if st == nil {
+		t.Fatal("R-218: the box went SILENT after recovering its key while still having no off-site target — the hub's staged credential is never collected and nothing ever asks again")
+	}
+	if st.State != OffsiteStateNeedsCredential {
+		t.Fatalf("declared state = %q, want %q", st.State, OffsiteStateNeedsCredential)
+	}
+}
+
+// SCENARIO E — and once the tier ACTUALLY WORKS the box goes quiet. This is the condition that
+// replaces the deleted one, and the pair above/below is what makes the deletion safe.
+func TestOffsiteDeclare_ConfiguredTierIsSilent(t *testing.T) {
+	m, sett := bareManager(t)
+	if err := sett.SetHubEscrowIdentityPresent(true); err != nil {
+		t.Fatal(err)
+	}
+	if err := sett.SetOffboxTarget(&settings.OffboxTarget{
+		Enabled: true, Host: "nas.local", Port: 22, User: "felhom", RepoPath: "/srv/repo",
+		Schedule: "daily", EscrowState: "escrowed",
+	}); err != nil {
+		t.Fatal(err)
+	}
+	st := m.OffboxReportStatus()
+	if st == nil {
+		t.Fatal("a configured tier must still report its ordinary off-site object")
+	}
+	if st.State == OffsiteStateNeedsCredential {
+		t.Fatal("a box whose tier is configured must not keep asking for a credential")
 	}
 }
 

controller/internal/report/escrow_confirm.go

@@ -20,6 +20,11 @@ type EscrowStatus struct {
 	IdentityBlobPresent bool   `json:"identity_blob_present"`
 	ResticPwSHA256      string `json:"restic_pw_sha256"`
 	CreatedAt           string `json:"created_at"`
+	// SupersededPresent / SupersededAt (v0.201.0, R-222) — the hub is ALSO keeping an earlier sealed
+	// package, and when it was set aside. Absent on a pre-0.97.0 hub, which reads as "no earlier
+	// package" and simply keeps today's message: an older hub cannot make the screen say anything new.
+	SupersededPresent bool   `json:"superseded_present"`
+	SupersededAt      string `json:"superseded_at"`
 }
 
 // EscrowAutoConfirmer runs the auto-confirm check on each report ACK. Long-lived (one per process) so
@@ -54,6 +59,12 @@ type EscrowAutoConfirmer struct {
 	// immediately, so the one fact that distinguishes a REBUILT box from a box that never had
 	// off-site backups was thrown away on every cycle. nil → not recorded (older wiring, tests).
 	RecordPresence func(present bool) error
+	// RecordSuperseded persists the ACK's `superseded_present` / `superseded_at` (v0.201.0, R-222) —
+	// whether the hub is ALSO keeping an EARLIER sealed package for this box. Wired here for the same
+	// reason as RecordPresence: this is the one place the ACK's escrow object already arrives, and a
+	// second wiring point in main.go is how this project accumulated six features that were built and
+	// never wired. nil → not recorded (older wiring, tests).
+	RecordSuperseded func(present bool, at string) error
 	Logger           *log.Logger
 
 	mu         sync.Mutex
@@ -111,6 +122,14 @@ func (c *EscrowAutoConfirmer) Reconcile(es *EscrowStatus) {
 			c.logf("[WARN] [escrow-confirm] could not record the hub's identity-blob presence (present=%v): %v", es.IdentityBlobPresent, err)
 		}
 	}
+	// Same discipline, same place, same reason (R-222): recorded unconditionally, because the box that
+	// needs it most is the unconfigured rebuilt one every gate below skips. Failure is logged, never
+	// swallowed, and never blocks the auto-confirm.
+	if c.RecordSuperseded != nil {
+		if err := c.RecordSuperseded(es.SupersededPresent, es.SupersededAt); err != nil {
+			c.logf("[WARN] [escrow-confirm] could not record the hub's superseded-package state (present=%v): %v", es.SupersededPresent, err)
+		}
+	}
 	c.mu.Lock()
 	c.sealedAt = es.CreatedAt // in-memory only; a timestamp, never a secret
 	c.mu.Unlock()

controller/internal/settings/settings.go

@@ -70,6 +70,22 @@ type Settings struct {
 	// first ACK — which is correct, because the hub is the authority on what the hub holds.
 	HubEscrowIdentityPresent bool `json:"hub_escrow_identity_present,omitempty"`
 
+	// HubEscrowSupersededPresent / HubEscrowSupersededAt (v0.201.0, R-222) cache the report ACK's
+	// `escrow.superseded_present` / `superseded_at` — whether the hub is ALSO keeping an EARLIER
+	// sealed package for this box, and when it was set aside.
+	//
+	// They exist for one message and nothing else. Without them the recovery screen cannot tell a
+	// genuinely wrong recovery code from a code that is RIGHT about an earlier package, because both
+	// look identical from here: the unseal fails closed against the package the hub currently serves.
+	// Measured on 2026-08-05 (CAMPAIGN-11 Phase 3 step 7) — the customer entered the correct code for
+	// their orphaned history and was told to check their typing.
+	//
+	// ⚠ These grant NO read path. The hub serves the CURRENT package only; retrieving a superseded one
+	// is a link that has never been built (R-199's inventory). The screen may say an earlier package is
+	// kept and when — both true, both non-secret — and must not promise it can be opened.
+	HubEscrowSupersededPresent bool   `json:"hub_escrow_superseded_present,omitempty"`
+	HubEscrowSupersededAt      string `json:"hub_escrow_superseded_at,omitempty"`
+
 	// RecoveryNoticePostponed (v0.200.0, R-193) — the customer chose "most nem" on the full-page
 	// recovery screen. It suppresses THE FULL-PAGE INTERRUPTION ONLY. The entry point in the backups
 	// area stays, permanently, for as long as the situation lasts: the data is still there whether or
@@ -647,6 +663,28 @@ func (s *Settings) SetHubEscrowIdentityPresent(present bool) error {
 	return s.save()
 }
 
+// GetHubEscrowSuperseded reports whether the hub is keeping an EARLIER sealed package for this box,
+// and when it was set aside (R-222). Both zero until an ACK has said so.
+func (s *Settings) GetHubEscrowSuperseded() (present bool, at string) {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+	return s.HubEscrowSupersededPresent, s.HubEscrowSupersededAt
+}
+
+// SetHubEscrowSuperseded records the ACK's `escrow.superseded_present` / `superseded_at`. Same
+// last-write-wins mirror discipline as SetHubEscrowIdentityPresent — the hub is the authority, and a
+// customer RESET that clears the retained rows must be able to turn the box's message back off.
+// Saves only on a change.
+func (s *Settings) SetHubEscrowSuperseded(present bool, at string) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	if s.HubEscrowSupersededPresent == present && s.HubEscrowSupersededAt == at {
+		return nil
+	}
+	s.HubEscrowSupersededPresent, s.HubEscrowSupersededAt = present, at
+	return s.save()
+}
+
 // ── Recovery screen (v0.200.0, R-193) ──────────────────────────────────────────
 
 // GetRecoveryNoticePostponed reports whether the customer chose "most nem" on the recovery page.

controller/internal/web/recovery_gate_test.go

@@ -0,0 +1,344 @@
+package web
+
+import (
+	"context"
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"gitea.dooplex.hu/admin/felhom-controller/internal/agentapi"
+	"gitea.dooplex.hu/admin/felhom-controller/internal/backup"
+	"gitea.dooplex.hu/admin/felhom-controller/internal/settings"
+)
+
+// CAMPAIGN-11 fixes — R-216, R-217, R-219, R-222, R-215.
+//
+// HANDLER-LEVEL throughout. Every assertion below is on the RENDERED PAGE or an on-disk effect,
+// because each of these defects lived in the handler while a helper-level test looked green: R-215's
+// predicate was correct and its page never asked, and R-217's failure branch built a struct whose
+// zero value the TEMPLATE misread. A test that stops at the helper observes neither.
+
+// postUnlockWith drives the real unlock handler with a code.
+func postUnlockWith(t *testing.T, s *Server, code string) *httptest.ResponseRecorder {
+	t.Helper()
+	form := url.Values{"recovery_code": {code}}
+	req := httptest.NewRequest(http.MethodPost, "/recovery/unlock", strings.NewReader(form.Encode()))
+	req.Header.Set("Content-Type", "application/x-www-form-urlencoded")
+	rr := httptest.NewRecorder()
+	s.recoveryUnlockHandler(rr, req)
+	return rr
+}
+
+// blamesTyping reports whether the page tells the customer to check how they typed their code. It is
+// the ONE sentence that may only ever appear for a genuinely wrong code.
+func blamesTyping(body string) bool {
+	return strings.Contains(body, "z szót pontosan") || strings.Contains(body, "nem fogadtuk el")
+}
+
+// ── SCENARIO A — an agent that cannot answer is NAMED, never the customer's code (R-216) ─────────
+//
+// THE HEADLINE. Measured live 2026-08-05: agent 0.120.0 404s the recovery route, the unlock was
+// attempted anyway, and a CORRECT code came back as "we did not accept your recovery code, check
+// that all ten words…" in 0.134 s.
+//
+// RED-PROOF: delete the `if support := s.recoverySupport(...)` block from recoveryUnlockHandler and
+// this test FAILS on the blamesTyping assertion — the accusation returns verbatim. Demonstrated
+// failing before this test was kept.
+func TestRecoveryGate_A_OldAgentIsNamed_NeverTheCode(t *testing.T) {
+	for _, tc := range []struct {
+		name  string
+		state agentapi.SupportState
+	}{
+		{"agent predates the route (404)", agentapi.SupportNo},
+		{"agent cannot be asked at all", agentapi.SupportUnknown}, // fail-CLOSED: §7.1
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			f := newRecoveryFixture(t)
+			f.s.SetRecoverySupport(func(context.Context) agentapi.SupportState { return tc.state })
+			// Model the REAL agent: one that cannot answer the route also fails the call. Without
+			// this the fake would succeed and the red-proof could not show the accusation returning —
+			// the mutation would fail the test for weaker reasons than the finding itself.
+			f.rec.fail = true
+
+			rr := postUnlockWith(t, f.s, testRecoveryCode)
+			if rr.Code != http.StatusOK {
+				t.Fatalf("unlock = %d", rr.Code)
+			}
+			body := rr.Body.String()
+
+			// EFFECT 1 — the customer is NOT blamed. This is the whole finding.
+			if blamesTyping(body) {
+				t.Fatal("R-216 RETURNED: an agent that cannot answer is still reported as a wrong recovery code")
+			}
+			// EFFECT 2 — the machine is named as the thing that cannot do it yet.
+			if !strings.Contains(body, "Ez a gép még nem tudja") {
+				t.Errorf("the refusal must say THE MACHINE cannot do this yet; got: %q", firstAlert(body))
+			}
+			// EFFECT 3 — it says the code is fine and was not consumed, so the customer keeps it.
+			if !strings.Contains(body, "kódoddal semmi baj") {
+				t.Error("the refusal must reassure the customer that their code is fine")
+			}
+			// EFFECT 4 — no attempt was made. An attempt that cannot succeed must never be made,
+			// because its failure is what got attributed to the code.
+			if n := len(f.rec.codes); n != 0 {
+				t.Fatalf("the handler attempted an unlock it could not complete (%d attempt(s))", n)
+			}
+		})
+	}
+}
+
+// The supported case still works — the gate must not refuse everything.
+func TestRecoveryGate_A_SupportedAgentProceeds(t *testing.T) {
+	f := newRecoveryFixture(t)
+	rr := postUnlockWith(t, f.s, testRecoveryCode)
+	if rr.Code != http.StatusOK {
+		t.Fatalf("unlock = %d", rr.Code)
+	}
+	if len(f.rec.codes) != 1 {
+		t.Fatalf("a supported agent must be asked exactly once, got %d", len(f.rec.codes))
+	}
+	if _, ok := f.mgr.OffboxRepoPasswordHash(); !ok {
+		t.Fatal("the recovered key was not placed")
+	}
+}
+
+// The declared coupling itself: the table must carry the row, at the version that ships the route.
+func TestRecoveryGate_A_CouplingIsDeclared(t *testing.T) {
+	if got := agentapi.MinAgentFor(agentapi.FeatureOffsiteKeyRecovery); got != "0.125.0" {
+		t.Fatalf("FeatureOffsiteKeyRecovery MinAgent = %q, want 0.125.0 (the version shipping POST /escrow/recover-offsite-password)", got)
+	}
+}
+
+// ── SCENARIO G — opened / empty / unreadable are THREE sentences (R-217) ─────────────────────────
+//
+// RED-PROOF: replace `s.renderRecoveryUnlockedNoList(w, r, msg)` with
+// `s.renderRecovery(w, r, msg, "", &backup.OffsiteInventory{})` and this test FAILS — the zero value's
+// Empty=false falls through to .InvUntagged and the page claims the store opened with content it
+// could not attribute. Demonstrated failing before this test was kept.
+func TestRecoveryGate_G_UnreadableStoreNeverClaimsToHaveOpened(t *testing.T) {
+	f := newRecoveryFixture(t)
+	// The tier is up, but reading the repository fails.
+	f.mgr.SetOffboxRunner(func(context.Context, []string, ...string) ([]byte, error) {
+		return nil, context.DeadlineExceeded
+	})
+
+	body := postUnlockWith(t, f.s, testRecoveryCode).Body.String()
+
+	if strings.Contains(body, "van benne tartalom") {
+		t.Fatal("R-217 RETURNED: an UNREADABLE store is reported as opened, with unattributable content")
+	}
+	if strings.Contains(body, "nincs benne egyetlen") {
+		t.Fatal("an unreadable store must not be reported as opened-and-empty either")
+	}
+	// It says the key is safe and what is pending — no claim about the contents.
+	if !strings.Contains(body, "kulcs visszakerült") {
+		t.Errorf("the customer must be told the key came back; got: %q", firstAlert(body))
+	}
+}
+
+// The three states must be DISTINGUISHABLE, which is the actual requirement.
+func TestRecoveryGate_G_ThreeDistinctStates(t *testing.T) {
+	// (1) opened, with apps.
+	f1 := newRecoveryFixture(t)
+	f1.runner.snapshots = []map[string]any{{"short_id": "aaa1111", "time": "2026-08-01T10:00:00Z", "tags": []string{"immich"}}}
+	b1 := postUnlockWith(t, f1.s, testRecoveryCode).Body.String()
+
+	// (2) opened, empty.
+	f2 := newRecoveryFixture(t)
+	f2.runner.snapshots = nil
+	b2 := postUnlockWith(t, f2.s, testRecoveryCode).Body.String()
+
+	// (3) could not be read.
+	f3 := newRecoveryFixture(t)
+	f3.mgr.SetOffboxRunner(func(context.Context, []string, ...string) ([]byte, error) {
+		return nil, context.DeadlineExceeded
+	})
+	b3 := postUnlockWith(t, f3.s, testRecoveryCode).Body.String()
+
+	if !strings.Contains(b1, "immich") {
+		t.Error("state 1 (opened, with apps) must list the app")
+	}
+	if !strings.Contains(b2, "nincs benne egyetlen") {
+		t.Error("state 2 (opened, empty) must say so explicitly")
+	}
+	if strings.Contains(b3, "nincs benne egyetlen") || strings.Contains(b3, "van benne tartalom") {
+		t.Error("state 3 (unreadable) must claim nothing about the contents")
+	}
+}
+
+// ── SCENARIO H — the retained earlier package is NAMED, not blamed on the customer (R-222) ───────
+func TestRecoveryGate_H_SupersededPackageIsNamed(t *testing.T) {
+	f := newRecoveryFixture(t)
+	f.rec.fail = true // the unseal fails against the CURRENT package — the real engine behaviour
+	if err := f.sett.SetHubEscrowSuperseded(true, "2026-08-05T15:03:14Z"); err != nil {
+		t.Fatal(err)
+	}
+
+	body := postUnlockWith(t, f.s, testRecoveryCode).Body.String()
+
+	if blamesTyping(body) {
+		t.Fatal("R-222 RETURNED: a code that is RIGHT about a retained earlier package is blamed on the customer's typing")
+	}
+	if !strings.Contains(body, "nem töröltük") {
+		t.Errorf("the customer must be told the earlier package is kept; got: %q", firstAlert(body))
+	}
+	if !strings.Contains(body, "2026-08-05T15:03:14Z") {
+		t.Error("when the hub knows WHEN the earlier package was set aside, say so")
+	}
+	// §7.6 — it may state the facts and must NOT promise the earlier package can be opened.
+	for _, forbidden := range []string{"vissza tudod állítani", "megnyithatod", "vissza fogod kapni"} {
+		if strings.Contains(body, forbidden) {
+			t.Errorf("the message promises the earlier package can be opened (%q) — the read path does not exist", forbidden)
+		}
+	}
+}
+
+// And with NO retained package the wrong-code message is unchanged — the only one mentioning typing.
+func TestRecoveryGate_H_WithoutASupersededPackageItIsStillTheWrongCodeMessage(t *testing.T) {
+	f := newRecoveryFixture(t)
+	f.rec.fail = true
+
+	body := postUnlockWith(t, f.s, testRecoveryCode).Body.String()
+	if !blamesTyping(body) {
+		t.Fatal("a genuinely wrong code must still get the wrong-code message")
+	}
+	if strings.Contains(body, "nem töröltük") {
+		t.Fatal("a box with no retained earlier package must not claim one exists")
+	}
+}
+
+// ── SCENARIO I — /recovery is not reachable by typing the path (R-215) ──────────────────────────
+//
+// RED-PROOF: delete the `if !s.recoveryOffer()` guard from recoveryPageHandler and this test FAILS —
+// a box that never had off-site backups renders "this machine was reinstalled, your off-site backups
+// are there". Demonstrated failing before this test was kept.
+func TestRecoveryGate_I_DirectGetRefusedOnABoxThatNeverHadBackups(t *testing.T) {
+	f := newRecoveryFixture(t)
+	// The defining fact: the hub holds nothing for this box.
+	if err := f.sett.SetHubEscrowIdentityPresent(false); err != nil {
+		t.Fatal(err)
+	}
+	if f.s.recoveryOffer() {
+		t.Fatal("fixture error: the box still reads as offered")
+	}
+
+	rr := getRecoveryPage(t, f.s)
+	if rr.Code == http.StatusOK && strings.Contains(rr.Body.String(), "jratelep") {
+		t.Fatal("R-215 RETURNED: a box that never had off-site backups renders the recovery story on a direct GET")
+	}
+	if rr.Code != http.StatusFound {
+		t.Errorf("a direct GET on a non-offered box should redirect away, got %d", rr.Code)
+	}
+}
+
+// A box that IS in the situation still reaches the page directly — including after "most nem",
+// which is what makes the permanent backups-area entry point work at all.
+func TestRecoveryGate_I_DirectGetStillWorksWhenOffered(t *testing.T) {
+	f := newRecoveryFixture(t)
+	if err := f.sett.SetRecoveryNoticePostponed(true); err != nil {
+		t.Fatal(err)
+	}
+	if f.s.recoveryInterrupts() {
+		t.Fatal("fixture error: postpone did not suppress the interruption")
+	}
+	rr := getRecoveryPage(t, f.s)
+	if rr.Code != http.StatusOK {
+		t.Fatalf("an offered box must still reach /recovery directly after a postpone, got %d", rr.Code)
+	}
+}
+
+// ── SCENARIO F — the unlock finishes the job: key placed, tier brought up, then the listing ──────
+func TestRecoveryGate_F_UnlockBringsTheTierUpBeforeListing(t *testing.T) {
+	f := newRecoveryFixture(t)
+	// The pristine rebuilt shape: no coordinates at all, so the listing is impossible until the tier
+	// comes up. Removing the target is what makes this the shape (a) the screen exists for.
+	if err := f.sett.SetOffboxTarget(nil); err != nil {
+		t.Fatal(err)
+	}
+	called := 0
+	f.s.SetRecoveryTierUp(func(context.Context) error {
+		called++
+		// Bringing the tier up = coordinates exist afterwards.
+		return f.sett.SetOffboxTarget(&settings.OffboxTarget{
+			Enabled: true, Host: "nas.local", Port: 22, User: "felhom", RepoPath: "/srv/repo",
+			Schedule: "daily", EscrowState: "escrowed",
+		})
+	})
+	f.runner.snapshots = []map[string]any{{"short_id": "bbb2222", "time": "2026-08-02T09:00:00Z", "tags": []string{"calibre-web"}}}
+
+	body := postUnlockWith(t, f.s, testRecoveryCode).Body.String()
+
+	if called != 1 {
+		t.Fatalf("the unlock must bring the tier up exactly once, got %d", called)
+	}
+	if !strings.Contains(body, "calibre-web") {
+		t.Fatalf("the listing the screen PROMISES did not render; got: %q", firstAlert(body))
+	}
+}
+
+// A tier that cannot come up yet is PENDING, not a failure, and never a wrong code.
+func TestRecoveryGate_F_TierNotUpYetSaysPendingNotFailed(t *testing.T) {
+	f := newRecoveryFixture(t)
+	if err := f.sett.SetOffboxTarget(nil); err != nil {
+		t.Fatal(err)
+	}
+	f.s.SetRecoveryTierUp(func(context.Context) error { return context.DeadlineExceeded })
+
+	body := postUnlockWith(t, f.s, testRecoveryCode).Body.String()
+
+	if blamesTyping(body) {
+		t.Fatal("a tier that has not come up yet must never read as a wrong code")
+	}
+	if !strings.Contains(body, "kapcsolódási adatait") {
+		t.Errorf("say what is pending — the connection details; got: %q", firstAlert(body))
+	}
+	// The key IS placed regardless: nothing failed.
+	if _, ok := f.mgr.OffboxRepoPasswordHash(); !ok {
+		t.Fatal("the recovered key must be placed even when the tier cannot come up yet")
+	}
+}
+
+// firstAlert extracts the first rendered alert body, for readable failure output.
+func firstAlert(body string) string {
+	i := strings.Index(body, "alert-error")
+	if i < 0 {
+		i = strings.Index(body, "alert-info")
+	}
+	if i < 0 {
+		return "(no alert rendered)"
+	}
+	seg := body[i:]
+	if j := strings.Index(seg, "</div>"); j > 0 {
+		seg = seg[:j]
+	}
+	if len(seg) > 400 {
+		seg = seg[:400]
+	}
+	return strings.TrimSpace(seg)
+}
+
+// dataDirHasNoCode asserts the recovery code never reached disk — the sweep, with a positive control
+// in the caller. Kept here because every path above handles a code.
+func dataDirHasNoCode(t *testing.T, dir, code string) {
+	t.Helper()
+	var hits []string
+	_ = filepath.Walk(dir, func(p string, info os.FileInfo, err error) error {
+		if err != nil || info == nil || info.IsDir() {
+			return nil
+		}
+		b, rerr := os.ReadFile(p)
+		if rerr == nil && strings.Contains(string(b), code) {
+			hits = append(hits, p)
+		}
+		return nil
+	})
+	if len(hits) > 0 {
+		t.Fatalf("the recovery code was written to disk: %v", hits)
+	}
+}
+
+var _ = backup.HashResticPassword

controller/internal/web/recovery_handlers.go

@@ -5,6 +5,7 @@ import (
 	"net/http"
 	"time"
 
+	"gitea.dooplex.hu/admin/felhom-controller/internal/agentapi"
 	"gitea.dooplex.hu/admin/felhom-controller/internal/backup"
 )
 
@@ -62,7 +63,23 @@ func recoveryNoStore(w http.ResponseWriter) {
 
 // recoveryPageHandler renders the screen (GET /recovery). Reachable whenever the situation holds —
 // including after "most nem", which is how the backups-area entry point can point at it forever.
+//
+// ⚠ GATED ON THE SAME PREDICATE AS THE INTERCEPTION (R-215, v0.201.0). It was not, and the page
+// rendered its whole story — *"Ezt a gépet újratelepítették. A korábbi, házon kívüli mentéseid
+// megvannak"* — on any claimed box, including one installed hours earlier that had never had an
+// off-site backup in its life. Measured on 2026-08-05 (CAMPAIGN-11 §F9): the hub held no escrow row
+// for that host at all, so `recoveryOffer()` was correctly false; the page simply never asked it.
+//
+// The predicate was right, the POST sibling below already consulted it, and the backups-area template
+// gates the identical sentence on `.RecoveryOffer` — this GET was the one door left open. A box with
+// no link to /recovery is only reached by typing the path or following a stale bookmark, but what it
+// then asserts is false in two ways at once, on the one screen whose purpose is to be believed about
+// backups.
 func (s *Server) recoveryPageHandler(w http.ResponseWriter, r *http.Request) {
+	if !s.recoveryOffer() {
+		http.Redirect(w, r, "/backups/remote", http.StatusFound)
+		return
+	}
 	s.renderRecovery(w, r, "", "", nil)
 }
 
@@ -70,6 +87,19 @@ func (s *Server) recoveryPageHandler(w http.ResponseWriter, r *http.Request) {
 // read-only listing). errorMsg/flash are already-customer-facing Hungarian; NEITHER EVER CONTAINS THE
 // CODE — see unlockHandler.
 func (s *Server) renderRecovery(w http.ResponseWriter, r *http.Request, errorMsg, flash string, inv *backup.OffsiteInventory) {
+	s.renderRecoveryState(w, r, errorMsg, flash, inv, false)
+}
+
+// renderRecoveryUnlockedNoList is the R-217 shape: the unlock SUCCEEDED but the repository could not
+// be listed. It renders the unlocked page with NO listing block at all — never the zero-value
+// inventory, whose Empty=false made the template claim the store had opened with unattributable
+// content. "Opened with apps", "opened and empty" and "could not be read" are three states and this
+// is the third.
+func (s *Server) renderRecoveryUnlockedNoList(w http.ResponseWriter, r *http.Request, errorMsg string) {
+	s.renderRecoveryState(w, r, errorMsg, "", nil, true)
+}
+
+func (s *Server) renderRecoveryState(w http.ResponseWriter, r *http.Request, errorMsg, flash string, inv *backup.OffsiteInventory, unlockedNoList bool) {
 	recoveryNoStore(w)
 	data := s.baseData("recovery", "Adatok visszaszerzése")
 	data["CSRFField"] = s.csrfField(r)
@@ -87,7 +117,13 @@ func (s *Server) renderRecovery(w http.ResponseWriter, r *http.Request, errorMsg
 		data["Unlocked"] = true
 		data["InvApps"] = inv.Apps
 		data["InvEmpty"] = inv.Empty
+		// InvUntagged is "the repository OPENED and holds snapshots we could not attribute to an app".
+		// It is only ever computed from a real reading — a failed read reaches the branch below and
+		// renders no listing at all (R-217).
 		data["InvUntagged"] = !inv.Empty && len(inv.Apps) == 0
+	} else if unlockedNoList {
+		data["Unlocked"] = true
+		data["InvUnavailable"] = true
 	}
 	w.Header().Set("Content-Type", "text/html; charset=utf-8")
 	if err := s.tmpl.ExecuteTemplate(w, "recovery", data); err != nil {
@@ -124,6 +160,48 @@ func (s *Server) SetRecoveryRecoverer(fn func() (backup.OffsiteKeyRecoverer, err
 	s.recoveryRecovererFn = fn
 }
 
+// recoverySupport is the R-216 capability verdict for the unlock path. It uses the SAME machinery the
+// NAS gate uses (`netFeatures` + the agent client as prober), so the version channel decides without
+// probe traffic whenever the agent's version is known.
+//
+// It returns SupportUnknown when the agent handle cannot be built — and the CALLER treats Unknown as
+// "cannot ask". That is deliberate and is the whole point: a box that cannot reach its agent must not
+// attempt an unlock whose failure would be read as a bad code.
+func (s *Server) recoverySupport(ctx context.Context) agentapi.SupportState {
+	if s.recoverySupportFn != nil {
+		return s.recoverySupportFn(ctx)
+	}
+	agent, err := s.agentClient()
+	if err != nil {
+		return agentapi.SupportUnknown
+	}
+	state, source := s.netFeatures.SupportsWithSource(ctx, agent, agentapi.FeatureOffsiteKeyRecovery)
+	s.logger.Printf("[DEBUG] [web] recovery capability gate: %s=%s (source=%s)", agentapi.FeatureOffsiteKeyRecovery, state, source)
+	return state
+}
+
+// SetRecoverySupport overrides the capability verdict (tests). INIT-ONLY.
+func (s *Server) SetRecoverySupport(fn func(context.Context) agentapi.SupportState) {
+	s.recoverySupportFn = fn
+}
+
+// recoverySuperseded reports the hub's statement that an EARLIER sealed package is kept, and when
+// (R-222). Both zero on a pre-0.97.0 hub, which keeps the old message — an older hub simply cannot
+// make the screen claim anything new.
+func (s *Server) recoverySuperseded() (bool, string) {
+	if s.settings == nil {
+		return false, ""
+	}
+	return s.settings.GetHubEscrowSuperseded()
+}
+
+// SetRecoveryTierUp wires the off-site apply-bridge reconcile (R-219 / R-218). INIT-ONLY.
+//
+// The unlock calls it between placing the recovered key and reading the repository — the step that
+// turns the screen's promised listing from structurally impossible into merely conditional. nil (or
+// an error) is survivable: the key is placed either way and the listing branch says what is pending.
+func (s *Server) SetRecoveryTierUp(fn func(context.Context) error) { s.recoveryTierUp = fn }
+
 // recoveryUnlockHandler takes the recovery code and drives the SHARED core (POST /recovery/unlock).
 //
 // R HANDLING, and it must not drift from the CLI's:
@@ -155,6 +233,23 @@ func (s *Server) recoveryUnlockHandler(w http.ResponseWriter, r *http.Request) {
 		s.renderRecovery(w, r, "A gép házon belüli kapcsolata most nem elérhető — próbáld újra néhány perc múlva.", "", nil)
 		return
 	}
+	// ── MESSAGE 2 of 4 — THE MACHINE CANNOT ASK (R-216). This gate FAILS CLOSED. ────────────────
+	//
+	// The unseal happens in the agent, behind POST /escrow/recover-offsite-password, which ships in
+	// agent v0.125.0. On an older agent that route 404s — and until v0.201.0 the unlock was attempted
+	// anyway and the 404 came back as "we did not accept your recovery code, check your ten words".
+	// A correct code, refused in 0.134 s, blamed on the customer (CAMPAIGN-11 Phase 1, the headline).
+	//
+	// So: anything other than SupportYes stops here. Unknown counts as cannot-ask, against this
+	// package's fail-open default — see FeatureOffsiteKeyRecovery for why this one feature inverts it.
+	// An attempt that cannot succeed must never be made, because its failure is attributed to the code.
+	if support := s.recoverySupport(r.Context()); support != agentapi.SupportYes {
+		code = ""
+		s.logger.Printf("[WARN] [web] recovery: refused before attempting — agent capability %s=%s (needs %s)",
+			agentapi.FeatureOffsiteKeyRecovery, support, agentapi.MinAgentFor(agentapi.FeatureOffsiteKeyRecovery))
+		s.renderRecovery(w, r, "Ez a gép még nem tudja megnyitni a mentéseidet — a hozzá tartozó házon belüli szolgáltatás régebbi, mint amit ehhez a lépéshez használunk. A kódoddal semmi baj, és nem is használtuk fel: tedd el biztonságos helyen. A gép magától frissül; próbáld újra később, vagy szólj a Felhom ügyfélszolgálatának, ha egy nap múlva sem működik.", "", nil)
+		return
+	}
 	ctx, cancel := context.WithTimeout(context.Background(), 90*time.Second)
 	defer cancel()
 	// THE SHARED CORE — the same function the command line drives. There is no second recovery
@@ -165,6 +260,31 @@ func (s *Server) recoveryUnlockHandler(w http.ResponseWriter, r *http.Request) {
 		// The agent's error names the STEP (fetch / unseal / place) and carries no secret. It is not
 		// shown raw: a customer needs to know what to check, not what age's KDF returned.
 		s.logger.Printf("[WARN] [web] recovery: unlock failed: %v", rerr)
+		// ── MESSAGE 4 of 4 — AN EARLIER PACKAGE IS KEPT (R-222) ────────────────────────────────
+		//
+		// The unseal failed against the package the hub CURRENTLY holds. That is the right outcome for
+		// a wrong code — and it is also the right outcome for a code that is perfectly correct about an
+		// EARLIER package the hub is deliberately keeping. From here the two are indistinguishable: the
+		// engine fails closed either way, which is exactly what it should do.
+		//
+		// So when the hub has told us an earlier package exists, say that instead of blaming typing.
+		// Measured on 2026-08-05 (CAMPAIGN-11 Phase 3 step 7): the customer entered the correct code
+		// for their orphaned history, got a real 1.111 s unseal attempt, and was told to check their
+		// ten words — while the same screen showed a seal date from after the code they were holding.
+		//
+		// ⚠ It states two facts the hub knows and STOPS. It does not promise the earlier package can be
+		// opened, because it cannot be: serving a superseded blob is an unbuilt link (R-199's
+		// inventory), and a conditional promise that turns out false on this screen is worse than
+		// saying less (the R-202 lesson).
+		if present, at := s.recoverySuperseded(); present {
+			when := ""
+			if at != "" {
+				when = " (" + at + ")"
+			}
+			s.renderRecovery(w, r, "Ez a kód nem nyitja meg azt a csomagot, amit most őrzünk ehhez a géphez. Ha egy korábbi kódot adtál meg: a géped azóta új mentési kulcsot kapott, és a régebbi csomagot"+when+" nem töröltük — megőrizzük. Megnyitni viszont innen egyelőre nem lehet, ezért ha a régebbi mentéseidre van szükséged, keresd a Felhom ügyfélszolgálatát. A kódoddal semmi nem történt, és semmi nem változott.", "", nil)
+			return
+		}
+		// ── MESSAGE 1 of 4 — THE CODE DID NOT OPEN IT. The ONLY one that mentions typing. ──────────
 		s.renderRecovery(w, r, "A megadott helyreállítási kódot nem fogadtuk el. Ellenőrizd, hogy mind a tíz szót pontosan, szóközökkel elválasztva írtad be — a kis- és nagybetűk nem számítanak. Semmi nem változott, nyugodtan próbáld újra.", "", nil)
 		return
 	}
@@ -175,21 +295,62 @@ func (s *Server) recoveryUnlockHandler(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 	s.logger.Printf("[INFO] [web] recovery: the offsite repository key was recovered and placed (outcome=%s)", res.Outcome)
+
+	// ── FINISH THE JOB (R-219, v0.201.0) ───────────────────────────────────────────────────────
+	//
+	// The screen promises: *"feloldjuk a mentéseid zárolását és megmutatjuk, mi van bennük"*. On the
+	// shape this screen exists for that promise was structurally unkeepable, and this call is what
+	// makes it possible at all:
+	//
+	//   - listing needs an off-site TARGET (OffsiteInventoryList returns errNoOffsiteTarget without);
+	//   - a target cannot exist without a repository password, because ApplyOffsiteTarget →
+	//     WriteOffboxSecrets MINTS one whenever none is present;
+	//   - shape (a) is DEFINED by having no repository password. So shape (a) ⇒ no target ⇒ the
+	//     listing could never render. Not sometimes. Ever.
+	//
+	// And there was no second chance: placing the key flips recoveryOffer() false (a password now
+	// exists and the box is not orphaned), so the unlock response was the customer's ONLY opportunity
+	// to see the listing — and it was guaranteed not to contain it.
+	//
+	// So bring the tier up here, synchronously, between placing the key and reading the repository.
+	// It is also the second half of R-218's deadlock: the apply-bridge only retried "on next config
+	// refresh/restart", and nothing triggered either.
+	if s.recoveryTierUp != nil {
+		tctx, tcancel := context.WithTimeout(context.Background(), 90*time.Second)
+		if terr := s.recoveryTierUp(tctx); terr != nil {
+			// NOT a failure of the unlock — the key IS placed. Logged, never swallowed, and the
+			// listing branch below says what is pending rather than what failed.
+			s.logger.Printf("[WARN] [web] recovery: the offsite tier could not be brought up yet: %v", terr)
+		}
+		tcancel()
+	}
+
 	// Unlocked. Now show what is in there — read-only.
 	ictx, icancel := context.WithTimeout(context.Background(), 2*time.Minute)
 	defer icancel()
 	inv, ierr := s.backupMgr.OffsiteInventoryList(ictx)
 	if ierr != nil {
+		// ── MESSAGE 3 of 4 — THE STORE COULD NOT BE READ (R-217) ────────────────────────────────
+		//
+		// ⚠ PASS nil, NOT backup.OffsiteInventory{}. The zero value has Empty=false and Apps=nil, which
+		// the template reads as InvUntagged — *"A tároló megnyílt, és van benne tartalom, de nem tudtuk
+		// alkalmazásokhoz rendelni."* Three assertions, none of them known: it did not open, its
+		// contents are unknown, and attribution is not the problem. Measured on 2026-08-05 (CAMPAIGN-11
+		// Phase 1), rendered directly beneath the honest message, contradicting it.
+		//
+		// OffsiteInventory.Empty exists precisely to prevent this — its own doc comment says it is
+		// "named rather than inferred from len(Apps)==0, WHICH IS ALSO WHAT A FAILED READ LOOKS LIKE".
+		// The field built for the hazard was defaulted past. nil means "no inventory to render", and
+		// the template shows no listing block at all.
 		s.logger.Printf("[WARN] [web] recovery: unlocked but the inventory could not be read: %v", ierr)
-		empty := backup.OffsiteInventory{}
-		msg := "A kulcs visszakerült, de a mentések listáját most nem sikerült beolvasni. Nézd meg a Biztonsági mentés oldalt néhány perc múlva."
+		msg := "A kulcs visszakerült, de a mentések listáját most nem sikerült beolvasni. A mentéseid nincsenek veszélyben — nézd meg a Biztonsági mentés oldalt néhány perc múlva."
 		if backup.ErrNoOffsiteTarget(ierr) {
-			// The pristine rebuilt shape: the key is in place but the box has no off-site coordinates
-			// yet. It resolves by itself once the tier is re-applied, so say that rather than showing a
-			// failure the customer cannot act on.
-			msg = "A kulcs visszakerült. A gép még most kapcsolódik újra a házon kívüli tárhelyhez — a mentéseid listája néhány perc múlva jelenik meg a Biztonsági mentés oldalon."
+			// The tier could not be brought up in time — the transport credential has not arrived yet.
+			// Nothing has failed; something is pending, and the declaration stays alive (R-218) so the
+			// hub's self-heal can still act.
+			msg = "A kulcs visszakerült, és biztonságban van. A gép még várja a házon kívüli tárhely kapcsolódási adatait — amint megvannak, a mentéseid listája megjelenik a Biztonsági mentés oldalon. Nincs teendőd."
 		}
-		s.renderRecovery(w, r, msg, "", &empty)
+		s.renderRecoveryUnlockedNoList(w, r, msg)
 		return
 	}
 	s.renderRecovery(w, r, "", "A mentéseid zárolása feloldva. Az alábbiakat találtuk a tárolóban — semmit nem állítottunk vissza.", &inv)

controller/internal/web/recovery_test.go

@@ -16,6 +16,7 @@ import (
 	"testing"
 	"time"
 
+	"gitea.dooplex.hu/admin/felhom-controller/internal/agentapi"
 	"gitea.dooplex.hu/admin/felhom-controller/internal/backup"
 	"gitea.dooplex.hu/admin/felhom-controller/internal/config"
 	"gitea.dooplex.hu/admin/felhom-controller/internal/settings"
@@ -127,6 +128,11 @@ func newRecoveryFixture(t *testing.T) *recoveryFixture {
 	s := &Server{cfg: cfg, settings: sett, backupMgr: mgr, stackMgr: stackMgr, logger: lg, version: "test"}
 	s.loadTemplates()
 	s.SetRecoveryRecoverer(func() (backup.OffsiteKeyRecoverer, error) { return rec, nil })
+	// R-216: the capability gate FAILS CLOSED, so the default fixture states the supported case
+	// explicitly. Without this every unlock test would exercise the refusal instead — which is exactly
+	// the protection working, and exactly not what those tests are about. The refusal has its own
+	// tests in recovery_gate_test.go, each overriding this.
+	s.SetRecoverySupport(func(context.Context) agentapi.SupportState { return agentapi.SupportYes })
 	return &recoveryFixture{s: s, mgr: mgr, sett: sett, rec: rec, runner: rr, dataDir: cfg.Paths.DataDir}
 }
 

controller/internal/web/server.go

@@ -104,6 +104,13 @@ type Server struct {
 	// recoveryRecovererFn is the recovery screen's agent seam (nil → the shared agentClient(), the
 	// same channel the CLI uses). Tests inject a fake so the HANDLER itself can be driven.
 	recoveryRecovererFn func() (backup.OffsiteKeyRecoverer, error)
+	// recoverySupportFn overrides the R-216 agent-capability verdict for the unlock path (tests).
+	// nil → the real gate over netFeatures. See recoverySupport: Unknown means CANNOT ASK here.
+	recoverySupportFn func(context.Context) agentapi.SupportState
+	// recoveryTierUp brings the off-site tier up between placing a recovered key and reading the
+	// repository (R-219). Wired from main.go to the apply-bridge's Reconcile; nil → skipped, and the
+	// listing branch then reports what is pending rather than claiming a failure.
+	recoveryTierUp func(context.Context) error
 
 	// NAS add orchestration (verify-before-commit): the single-flight job slot + the two seams.
 	// netAgentFn nil → the shared agentClient(); netProbeFn nil → runNetProbe (the uid-1000 re-exec).

controller/internal/web/templates/recovery.html

@@ -18,7 +18,12 @@
     {{if .Flash}}<div class="alert alert-info">{{.Flash}}</div>{{end}}
     {{if .Error}}<div class="alert alert-error">{{.Error}}</div>{{end}}
 
-    {{if .InvEmpty}}
+    {{if .InvUnavailable}}
+    <!-- R-217: the unlock SUCCEEDED but the repository could not be read. It renders NO listing and
+         claims nothing about the contents — the .Error above already says what is pending. This
+         branch exists because the previous code passed a zero-value OffsiteInventory here, whose
+         Empty=false fell through to .InvUntagged and asserted the store had opened with content. -->
+    {{else if .InvEmpty}}
     <div class="alert alert-warning">
         A tároló megnyílt, de <strong>nincs benne egyetlen mentés sem</strong>. Ez azt jelenti, hogy a
         kulcs jó volt, de ehhez a géphez nem tartozik korábbi mentés. Ha korábban biztosan készültek