Reported from a real phone. Both were mine, both shipped in the switch.
1. THE PHONE MENU WAS EMPTY. The drawer slid open correctly but had no items in
it. site.css's OLD mobile nav was still reaching the shell's sidebar: under
@media (max-width: 900px) it sets .nav-links to position:fixed; right:-100%,
which parked the menu off-canvas INSIDE the drawer. Measured at 390px and
860px: position fixed, right -280px.
The cause is a hole in my own check. At STOP 1 I diffed the two stylesheets'
declaration sets and found "exactly six" leaking declarations - but that diff
EXCLUDED @media blocks. Re-run media-aware, the real number is FORTY, and the
worst of them only apply below 900px, which is why desktop looked right and a
phone did not. A check that silently skips part of its subject is not a check;
this one reported a confident, wrong number.
Fixed at the source rather than per-property: every selector belonging to the
website's OLD chrome (nav, .nav-links*, .mobile-overlay, .hamburger, .logo*,
.lang-globe-li - 29 occurrences) is now scoped `body:not(.shell)`. Scoped, not
deleted, so a future non-shell page still works. .hamburger-box and
.hamburger-line are deliberately left unscoped: the shell's burger uses them.
Leaks after the fix: 40 -> 0.
2. THE FOOTER RENDERED AS A SECOND COLUMN at the top right on 16 pages. body is a
flex row on a shell page, and my insert only moved the footer inside <main>
when the page ENDED with </main>; pages with scripts after it fell through and
left the footer as a direct child of body, i.e. a flex sibling of the sidebar.
Moved inside <main> on all 16, each move asserted to have landed inside.
Both are now gated: gate 3b requires every shell page's footer to be INSIDE
<main class="content">, red-proofed by putting one outside and watching it convict.
site.css?v=23 / shell.css?v=23.
The decoys plant a known defect and assert the gate convicts, so when the markup
they anchor on changes they must be rebuilt rather than deleted - which is exactly
what test_gate_decoys.py refused to let pass ('the decoy's anchor is gone or
ambiguous').
en-nav-drift drifts a SIDEBAR entry now, and expects gate 3b's message
rather than gate 3's - the website's <nav> is gone
leftover-text-switch anchors on <details class=lang-globe>; the old nav's
<li class=lang-globe-li> wrapper no longer exists
lang-attr <html lang=en class=no-js>, the progressive-enhancement hook
ascii-hungarian-text anchors on the new English home's own link text; its old
'Contact form' link went when the page became a real twin
The other decoys in that section still anchor correctly and were left alone -
checked each one against the new files rather than rewriting the lot.
Part D. Every public page now wears the controller's frame around the website's
own content. One commit, as asked: never half old, half new.
20 pages switched (12 Hungarian incl. the hidden prices page, 8 English)
website/preview/ deleted entirely - A, M1, M2 and M3 studies all gone
robots.txt: the Disallow: /preview/ line removed with them
site.css?v=22 / shell.css?v=22 on every page
PROVED before pushing, not after:
* HEAD TAGS: title, meta description, keywords, author, robots, canonical, every
hreflang, all seven Open Graph tags, all four Twitter tags and the JSON-LD
captured on all 19 pages before the switch and compared after. The ONLY
difference anywhere is the dash substitution inside the two FAQ pages'
FAQPage JSON-LD - verified by applying the same substitution to the "before"
copy and getting an exact match. Nothing else moved.
* VISIBLE TEXT: identical to the live pages word for word, legal texts identical
with no substitution permitted at all.
* No page is left noindex (the two download pages' noindex,follow is their own,
pre-existing and unchanged), no page mentions /preview/, every page has a
footer, no page overflows at 390px.
THE HEAD-TAG DIFF EARNED ITS PLACE. The home page was about to ship carrying
`noindex, nofollow` from the preview, and had lost meta keywords, meta author, all
seven Open Graph tags, all four Twitter tags and the LocalBusiness JSON-LD -
because its head was hand-built while every other page carried its live one over.
The diff caught it; nothing on the rendered page would have. The generator now
uses the live head for EVERY page including the home one.
szolgaltatasok-nonpublic.html was converted too. It is in no menu and was not in
the brief's list, but leaving one page on the old chrome would have left the site
half-switched - which is the one thing part D was not allowed to do.
scripts/site_gates.py: SHELL is now the real page set; MOCKUPS is empty; gate 3's
old nav/footer loop keeps working for any page added later WITHOUT the shell, and
skips the ones that have it.
Caught BEFORE the switch, not after. Stripping the website's chrome also stripped
<footer>, and the shell has none of its own. Every one of the 19 pages had lost:
the link to the Adatkezelesi tajekoztato and the Feltetelek, the operator's name,
info@felhom.eu, and the copyright line. The legal pages are in no menu, so they
would have become unreachable from navigation entirely.
The footer is now carried VERBATIM from each live page into the bottom of the
content column. Verified on all 19: footer present, legal link present, e-mail
present, and the footer sits INSIDE <main class=content> rather than becoming a
flex sibling of the sidebar.
Gate 3b now checks the footer as well as the sidebar across the shell set, so the
guarantee gate 3 used to give is not lost in the move. Red-proofed: removing one
page's footer convicts.
Staged under /preview/m3/. Nothing live. 19 shell pages in total with the home.
Each page keeps its own <head> byte-for-byte apart from two additions (shell.css
and the no-js/js hook), so title, meta description, canonical, hreflang, Open Graph
and JSON-LD all survive the move untouched. The body loses the website's
<nav>/<footer> and the old hamburger script, and gains the shell.
PROVED, not assumed - a word-level diff of every page's visible text, live vs
shell, with the nav and footer removed from both:
* 18 of 18 pages: visible text IDENTICAL after the one allowed substitution
* adatkezeles and feltetelek: identical with NO substitution permitted at all.
The legal texts are reviewed documents; the frame changed, the words did not.
That check earned itself immediately. The first version of the page-header
transform rebuilt the block from its <h1> and first <p> and SILENTLY DROPPED the
rest: the "Fontos tudnivalok" / "Good to know" disclaimer banner on both apps
pages and the 404's English line - 168 words across three pages. Nothing looked
wrong on the page. The diff found it.
Em dashes: 318 replaced with the Hungarian gondolatjel across 16 pages (gyik and
en/faq 104 each). The FAQ pages' JSON-LD is swept with the SAME substitution, so
gate 17's equality with the visible answers still holds - it does, on both.
scripts/site_gates.py:
* SHELL now holds all 19 pages.
* is_en(): a page is English if it is a TWINS value OR lives under an en/
directory. Without the second half the dash-language, <html lang> and
no-Hungarian gates all judged preview/m3/en/* as Hungarian - 30 false fails.
* norm_block() also drops `is-open` and `aria-expanded`, because the shell
renders the nav group containing the active page open, exactly as the
controller does, and that legitimately differs per page.
The last change LOOSENS a check, so it was red-proofed BOTH ways: a one-character
difference in one menu label still convicts, and a state-only difference (the
active marker moving to another item) correctly does not.
One generator now owns the shell. The home page's content comes from m3.py and its
shell from the same function every other page uses, so the two cannot drift - which
they already had, and gate 3b caught it.
The direction the operator chose: the controller's FRAME around the website's own
content. Staged at /preview/m3/, noindex, nothing live.
website/assets/shell.css - 117 rules COPIED VERBATIM from felhom-controller
style.css @ dcd89364a4 (sidebar, mobile top bar, off-canvas drawer, and the
component atoms the restyled pages use: btn, alert, app-info-card, launch-*,
stack-detail-card, form-*). Exactly ONE rule is not verbatim and says so:
body.shell { display:flex; min-height:100vh }, which lives in the controller's
base rather than in a component, scoped to .shell so other pages are untouched.
The generator NAMES every rule it copies and REFUSES to write the file if one has
disappeared upstream. That is not ceremony: it already caught two things this
session. It refused because .sidebar::-webkit-scrollbar does not exist (only the
-track override does), and because prefers-reduced-motion is split across FIVE
separate blocks in the controller - reading only the first would have silently
lost four rules. The M1 round lost a rule exactly this way.
The bare `body { flex-direction: column }` that broke M1's phone layout is copied
this time, by name, as are body.nav-open and the no-js/js pair.
Load order is deliberate: shell.css AFTER site.css. Thirteen selectors exist in
both files with different meanings (.nav-links, .page-header, .lang-globe*, .ico)
- a horizontal site nav there, a vertical sidebar here. Later wins, and a page
that does not load shell.css is completely unaffected.
website/preview/m3/index.html - the home page in the shell. Content is /preview/'s
cleaned text; body carries `shell page-index` so the existing .page-index content
rules (caps, flow, vs, band, who-mini) are reused rather than duplicated. Six
sections, left-aligned headers, the controller's cards. Head tags (title, meta
description, canonical, hreflang) are already the FINAL values, so the switch in
part D needs no head edit.
Four changes from the operator's mockup, each with its reason in the generator:
1. the active item follows the page (the mockups show Inditopult active on all)
2. Teszteles is ALSO a primary button at the top of the lower group - it is the
one action we want and the mockup buries it at the bottom
3. the Alkalmazasok categories are anchors on /alkalmazasok: real URLs, no
JS-only state
4. ADDED GYIK. It is in today's nav; dropping it silently would take a page out
of navigation, which is not a change a restyle gets to make.
scripts/site_gates.py - gate 3 extended rather than duplicated, as asked: a SHELL
set whose pages must carry an IDENTICAL <nav class="sidebar"> after the active
marker is normalised. RED-PROOFED: a one-character change to one menu label on a
second shell page convicts. It also prints "DIFFER" rather than a reassuring
"identical" count when it fails - a false reassurance printed beside a conviction
is the exact defect this repo files rows about.
documentation/design/design-system.md - new "Shell" section: what is shared (tokens
+ shell + atoms), what each side owns, how a change flows (controller first, the
website's copy the same week), and that the cross-repo gate is deliberately NOT
built yet.
Design studies only. Nothing live changes, nothing is linked from a live page, both
carry noindex and /preview/ is already Disallow-ed in robots.txt. felhom-controller
was read ONLY (baseline dcd89364a4); nothing in it was touched.
M1 website/preview/m1/ - the website wearing the controller's shell
index (Indítópult) welcome card + all 58 launcher tiles
alkalmazasok the catalog in the site's own five buckets, vanilla search
mentesek the three backup levels, every number marked "minta"
hogyan / teszteles / kapcsolat today's copy inside the dashboard shell
M2 website/preview/m2/ - short hero beside a window holding a clickable launcher
Real data, not invented:
* 58 apps from app-catalog-felhom.eu @ 8ea72aa: real names, logos, Hungarian
taglines, use cases and first steps out of each .felhom.yml app_info.
* Tile colours reproduce the controller's OWN tileColor (funcmap.go): FNV-1a 32
of the slug -> hue, 45% 38%. No brand_color exists in the catalog, so every
tile here is the colour a real launcher would paint.
* Page copy is today's live website only.
* Category labels are the five buckets alkalmazasok.html already uses. None invented.
The 58 is now measured, not quoted: 60 templates minus plant-it (abandoned) and
wger (hidden) = 58 offered, which is exactly what the site's prose says. See the
R-929 note in the audit folder.
Mockup rules honoured:
* No value pretends to be live - every sample wears a "minta" badge.
* No control does nothing: the disabled sidebar entries (Tárhely, Megosztás,
Rendszermonitor, Beállítások) say "a saját dobozodon működik" rather than
404-ing, and the contact form's submit is disabled with the same kind of hint.
The consent text is the live one, unchanged.
* App panels open with :target, so they work with JavaScript switched OFF and
their text is real HTML a crawler can read - not JS-built content.
* No login form anywhere. No external request, no AI or stock image.
CSS: website/preview/mockup.css is three labelled blocks so the shared-design
report can count them - adapted base (font URLs repointed /static -> /assets),
188 rules copied VERBATIM from the controller, and the mockup-only rules. The
burger is the website's own three-span control, copied rather than hand-rolling an
SVG: the shared sprite carries no i-menu.
scripts/site_gates.py gains a MOCKUPS set, with the reasoning in a comment. The
mockups keep every correctness check (BOM, emoji, banned strings and CDN hosts, no
<style>, cache-busting, <html lang>, clean tail, dashboard-picture language, the
picture-viewer rules) and are excluded from exactly the three chrome-consistency
checks a dashboard-shell study must fail by construction: nav/footer identical to
index.html, the NO_TWIN language globe, and analytics - the last deliberately, so a
hidden mockup cannot pollute the real stats. Delete the set with the directories.
site_gates.py green.
Part C. The live home page is NOT touched: every CSS rule added is a new class
under .page-index, so index.html renders byte-for-byte as before.
website/preview/index.html (BOM + CRLF, noindex, Disallow: /preview/)
Ten sections, ten different layout families. The three repeats the audit found
are gone: why-grid ran 3x and apps-showcase 2x.
- hero: a real dashboard screenshot instead of the logo, sized by its content
instead of a forced 100vh band, no infinite float animation, and a sentence
made of four facts instead of "Professzionalis ... telepites es uzemeltetes".
- the ten capability cards become three labelled clusters. Their TEXT is
verbatim: the audit's own finding was that the writing is right and the
layout is wrong, so only the layout moved. Every hedge ("ha van") survives.
- NEW "Ki all mogotte" with the operator's photo and five approved sentences.
- the 3 cloud-problem cards + the 4 own-server cards merge into ONE two-column
comparison. No new claim about any competitor: all four left-hand sentences
already existed on the page.
- the two tile walls (10 + 7 tiles duplicating pages that already exist) shrink
to a two-item band that keeps the app names and the links.
- section headers are left-aligned and varied, not 11x centred-two-words-plus-
one-blue-word.
Em dashes in the page's own copy: 18 -> 0. Deliberately kept: the 6 in <title>,
meta description, OG, Twitter and JSON-LD (operator decision at STOP 1 - indexed
strings deserve their own pass), and 1 in the shared footer, which gate 3 requires
to be identical across all 14 pages. Both stated, neither swept silently.
JetBrains Mono is no longer fetched. Exactly 13 characters were loading 31 KB:
the 1..5 process badges, the 01..05 service numbers, and - the audit missed this
at first and it is corrected in the file - the 3/2/1 of the backup rule.
assets/operator-portrait.webp 720x900, 93 KB
Cropped from an original that stays OUTSIDE this public repo. All metadata
stripped and VERIFIED BY READ-BACK: 0 EXIF tags, 0 GPS IFD entries, and none of
EXIF/XMP/ICCP/samsung/SM-A705FN/2023:10:18 appear anywhere in the bytes. Built by
re-writing the pixels into a fresh image, so the source info dict cannot travel.
Honest correction to the brief's premise: this photo had NO GPS tag to begin with.
scripts/site_gates.py
preview/index.html registered in NO_TWIN, with the reason. It cannot go in TWINS:
gate 3 wants its nav identical to index.html's (globe -> / and /en/) while the
twin check wants a pair's globe pointing at the pair's own URLs. The English twin
is written in the same commit as the switch, as a real twin. Everything else still
runs on the preview - BOM, emoji, nav/footer, globe, analytics, CDN, tokens,
cache-busting, dash-language, viewer, tail.
site_gates.py green. The builder (kept in the session scratchpad) writes BOM + CRLF
and asserts both: it first wrote LF and turned the nav/footer gate red for the
preview, which is exactly the trap this CHANGELOG's morning entry records.
COPY.md section 7: five Hungarian posts for Tue/Thu 19:00 over three weeks, each
with a source comment per claim and one link. The Monday post (2026-10-12, 6.2)
is untouched and no slot shares its day.
fb_probe.py gains --link (default unchanged). Every scheduled post used to
attach the home page's preview card whatever its text said, so a post about
/teszteles would have sent the people who clicked its card to the home page.
build_parser() split out of main() so a test can parse argv. Red-proved: both
new tests fail with "Namespace object has no attribute link" without the flag.
Search Console audit (seo-2026-10-10/not-indexed.md): all seven not-indexed
URLs with a verdict each - four expected by design, two waiting on Google's
crawl queue, one fault on our side. That one is /api/contact, the contact
form's POST endpoint crawled as a page; robots.txt now disallows /api/. The
performance baseline (0 clicks, 47 impressions, position 73, nine queries) is
recorded there to compare against in a month.
New page pair /teszteles + /en/testing: who the closed test is for, what a
tester needs (machine + internet quoted verbatim from the FAQ, plus a domain),
what they get (terms section 3, nothing new), what we ask for, and how many
places there are. Every "Jelentkezem tesztelonek" button now goes there.
Operator rulings 2026-10-10, recorded in CONTEXT.md:
D1 - the tester's domain is in the TESTER'S OWN name; Felhom pays its fee for
two years and the tester keeps it when the test ends or they leave; DNS
is managed with a token scoped to that one zone only.
D2 - the closed test runs with at most 5 households.
Terms 1.1 adds section 4 (the domain) with the D1 text; old sections 4-11 are
renumbered 5-12 and nothing else changed. Section 11 promises the testers an
e-mail before a change, so one is drafted for the operator to send.
Registered in site_gates.py TWINS/URL and listed in sitemap.xml.
The catalogue side is app-catalog-felhom.eu b7f0f7c. Here: the evidence, the
website, the register and the operator's view.
Website
- Two cards in the Otthon & Eletmod / Home & Lifestyle section of BOTH apps
pages, with assets: grocy-logo.svg is grocy's own icon with its single fill
made white like the other logos, lubelogger-logo.png is the app's own icon
with its dark background dropped and the mark made white (the rule
SparkyFitness's PNG follows), and six screenshots of each app's own UI with a
household's own data, taken headless on the bench from the published template.
- The app count moved 56 -> 58 in 15 places per language set, both languages,
and the open-source tile 49 -> 51. Checked by asking the same patterns for the
new number afterwards. marketing/facebook/COPY.md still says 56 and is NOT
changed: the post it carries is already scheduled.
Evidence
- documentation/audits/new-apps-2026-10-10/ — FIT.md (checklist group 0 for all
three, with the Hungarian-UI column), the bench and box transcripts, the
memory samples, the screenshots and the gate runs.
Register: 137 -> 139 rows, 2 opened, 0 closed.
- R-926 after a remove-keeping-backups and restore, nobody has checked what the
app page shows for an after_install app's generated password. Measured here:
LubeLogger is safe (its login is derived from the environment at every start,
the new password signs in, the data is back); grocy's install password still
signs in from the restored database but the deployed environment no longer
carries ADMIN_PASSWORD at all. Seven apps are in the class.
- R-927 Monica, stopped at checklist 0.2 with the measurements, waiting on the
operator.
Gate scripts: the same console trap in nineteen of them and in repo_gates.py
itself, where it ABORTED THE WHOLE RUNNER at the first gate — printing a
non-ASCII character on this workstation's cp1250 console raised
UnicodeEncodeError before the gate had decided anything, and reuse_refs_check.py
died while printing a NOTE. All now reconfigure their own streams. Red-proof
that the remaining script-test failures are not mine: test_due_checks_gate.py
fails the same 5 of 42 with the change reverted.
scripts/test_repo_gates_docstring.py pins the numbered list in repo_gates.py's
docstring against the GATES table IN ORDER. I listed poster-facts as 17 but
inserted it before decoy-coverage, so the sets matched and the order did not:
'same set, different order'. Caught on DooPlex, where script-tests runs (it
needs fcntl and cannot run on Windows).
Moved the table entry after decoy-coverage rather than renumbering the
docstring: the gate genuinely belongs last, since it only reports.
PART A -- the poster. documentation/architecture/felhom-system-poster.html
(307 KB). Secret scan first: ZERO IPv4, zero PEM blocks, zero ssh keys, zero
Bearer. The one EAA... match is base64 inside an embedded "mime":"font/woff2"
blob, not a Facebook token. "token"/"secret"/"password" appear 11 times and
every one is a NAME ("6. ep0 read token", "the hub seal key"); the poster
itself says "Names only; no secret values". All five long base64 blobs are
declared assets: 1 image/png, 3 text/javascript, 1 font/woff2.
It renders with NO network: the source mentions cdn.jsdelivr.net and Google
Fonts, but the loaded requests are only the HTML plus blob:/data: URLs -- the
bundler inlined everything. Measured, not assumed, and it matters: this is a
disaster-recovery document, so needing the internet to draw would be a defect.
No console errors.
The operator's three Claude Design fixes are all present: (a) no "WG" badge,
WireGuard only for the tunnel, no badge on the ep0-copy tile; (b) the box ->
ep0 arrow reads "encrypted on the box, sent through WireGuard"; (c) "Known
gaps" holds two items and NOT the household-keys sentence, which is now a
neutral "By design" note under the ep0 household namespace.
ONE FACT ON IT WAS WRONG. The felhom.eu tile said "served from DooPlex through
Cloudflare". It is not: Cloudflare is DNS only and the traffic goes direct --
measured this morning for the privacy notice, which states exactly that. The
poster would have contradicted a published page. Fixed in place (a label):
"served from DooPlex, Cloudflare DNS only". The first wording overflowed the
fixed-size tile, so it was shortened to fit and the evidence lives in the
facts file instead -- checked by re-rendering, not by hoping.
PART B -- the facts and the rule. DESIGN-PROMPT-...md is renamed
felhom-system-poster.facts.md (one home per fact), with the three fixes folded
in as explicit instructions so a regeneration cannot undo them, plus a new
"Badges" section saying a "WG" chip must never come back.
New rule, section 6 "The system poster stays true", added IDENTICALLY to all
five copies of unprompted-work.md (the four repos and the workspace root on
DooPlex; verified identical by diff before and after) and to
PROMPT-TEMPLATE.md's end-of-session checklist as a FIFTH coupled artifact.
scripts/poster_facts_gate.py WARNS when the facts file has a newer commit than
the poster. It never fails a push, deliberately: a refresh needs Claude Design
and the operator, --no-verify is forbidden here, so a blocking gate would leave
deleting it as the only way out. It compares COMMIT times, not mtimes, because
a checkout rewrites mtimes and every fresh clone would shout.
RED-PROOF -- and it found a real bug in the gate. The first run warned
correctly but exited 1: a single non-ASCII character in its own warning raised
UnicodeEncodeError on this cp1250 console. A gate whose entire contract is
"never fails a push" was failing pushes. Fixed (ASCII output + an encode
guard), and the decoy now asserts BOTH the warning and exit 0. Three branches
proven: facts newer -> warns, rc 0; poster newer -> quiet, rc 0; poster
missing -> "could not tell", rc 2, not a false all-clear.
The decoy itself was seen to fail, twice, on Linux (the suite needs fcntl and
cannot run on Windows): breaking the warning gives STALE_WARNS=False, and
making it exit 1 gives RC_STALE=1. All 80 felhom.eu decoys behave.
PART C -- do box reports pass through Cloudflare? NO. Two channels. DNS from
PUBLIC resolvers (not DooPlex's own, which answers the LAN address):
hub.felhom.eu is a CNAME to dooplex.hopto.org -> 37.191.56.193, not a
Cloudflare address, and no cf-ray comes back. The manifest: an ordinary k3s
Ingress, Cloudflare named only in a DNS setup comment. THE CONTROL that makes
the negative mean something: iso.felhom.eu resolves to 172.67.x / 104.21.x,
real Cloudflare addresses -- so the method does detect proxying.
So nothing is added to the Cloudflare row: the hub path does not touch it.
06-offsite-connectivity.md section 1 claimed the public edge is a
Cloudflare-Tunnel and "DooPlex has no public IP" -- both untrue today. Kept
and marked STALE with the measurement rather than rewritten, because that
paragraph is the reason ep0 exists and the argument needs its premise visible.
total-loss-of-dooplex.md's "today a CNAME to dooplex.hopto.org" is confirmed
correct.
Register: 137 before, 137 after, 0 opened, 0 closed -- every finding here was
small and fixed in the session.
DRAFTS. COPY.md section 6: two versions of the Page's first post, shortened
from section 2. Hungarian, tegezo, no price. 6.1 = 347 characters, 6.2 = 638
(counted as Unicode characters). Every claim carries a source comment naming
the line of website/index.html it rests on; the first line of each carries the
point alone, because Facebook cuts after about three lines.
Deliberate: ZERO emoji and ZERO hashtags, though the brief allows two of each.
The Felhom design system uses no emoji (the website gate holds it at 0) and
two hashtags would serve no real search.
CHECKED, because the post repeats it: "56 alkalmazas" is CORRECT. The apps
page carries 57 <div class="app-card"> but states 56, which looks off by one
until you read the category line -- "6 alkalmazas + 1 beepitett". The 57th
card is FileBrowser, built into every box and deliberately not counted.
index.html says "56 telepitheto alkalmazas" too. I nearly "fixed" a live page
into being wrong. NOT-A-FINDING.
SCHEDULE-POST. A third sub-command on fb_probe.py, reusing its token loader
(R-453), redaction, Bearer call and evidence writer:
- the body is READ FROM COPY.md by section name. The Hungarian never passes
through a shell or an argv string (brief 9.6); the caller names a section.
- published is ALWAYS "false" and NO argument can change it (brief 9.7).
The operator's review in Planner is the safety net, so an immediate post
must be unreachable, not merely not-the-default.
- check_when refuses a time under Meta's 10-minute floor or over its
6-month ceiling, BEFORE the call, so a bad time is a readable local
refusal rather than a Graph error.
- budapest_to_epoch uses the real tz database. If zoneinfo has no
Europe/Budapest it REFUSES rather than falling back to a hardcoded
+01:00/+02:00 -- guessing the offset is how a post goes out an hour wrong
across a DST boundary.
- list_scheduled tries /scheduled_posts then feed?is_published=false and
RECORDS WHICH ANSWERED; when both are refused it returns None so the
caller says "unproven" instead of claiming a removal it never saw.
TESTS: 30, of which 2 skip on Windows (no tzdata in this interpreter; the
command runs on the Linux host, which has the system zoneinfo).
RED-PROOF of the guard that matters, as the brief requires. Made
schedule_form accept published=..., ran ScheduleForm, and watched
test_no_argument_can_publish_immediately FAIL:
AssertionError: 'true' != 'false' : published changed published
Guard restored, all 30 green again.
No post has been made. The dry check and the real post come next; the
operator picks the version and the time first.
WHY .gitignore "was not working": it was working. git never consults
.gitignore for a file it ALREADY TRACKS. The rule `*secret*` matched fine --
proved by dropping an untracked copy in and watching check-ignore name
`.gitignore:3:*secret*`. The file had been tracked since feea0606, which is
ironically the commit that de-gitted the Resend key.
WHAT THE EXPOSED VALUE ACTUALLY WAS. Not an analytics password: the GITEA
ADMIN ACCOUNT PASSWORD (is_admin true; /api/v1/admin/users answered 200), in
a repo gitea.dooplex.hu serves anonymously to the internet. That is push
access to every repo -- including the one whose website/ is git-synced live
and whose scripts/ is published by tag to every new box installer (R-110).
Re-ranked P2 -> P1 on that measurement; my first ranking had only measured
the analytics blast radius.
Every committed value was still live. Nothing had ever been rotated.
ROTATED (values never echoed; written to a 0600 file on DooPlex):
umami-config APP_SECRET + POSTGRES_PASSWORD. The password was
changed INSIDE postgres (ALTER USER) as well as in the
Secret -- the env var is only read at first init, so
patching the Secret alone would have changed nothing.
healthchecks-config SECRET_KEY + SUPERUSER_PASSWORD (nothing consumes them,
there is no healthchecks Deployment).
gitea-creds no longer holds the admin password at all: a SCOPED
token (read:package + read:repository).
gitea admin new random password; gitea-system/gitea-admin updated.
VERIFIED, not assumed:
- new admin password -> 200, OLD PUBLISHED PASSWORD -> 401 (the leak is dead)
- umami: a real beacon returns 200 (so the app authenticates to postgres and
writes) while a bogus site id still returns 400 (so the 200 means something)
- hub: "Registry version check: latest = 0.304.0" AND "Template fetched
(5881 bytes)", no auth failures
- BOTH token scopes are load-bearing, and the second was found by breaking
it: a package-only token made the hub log "Template fetch: unexpected
status 403", because the template fetcher reads a raw file out of the
felhom-controller repo, not the registry.
AN INCIDENT CAUSED BY THE FIX, recorded because it is the useful part: the
rollout restart needed to pick up the new umami secret put umami into
CrashLoopBackOff and took stats.felhom.eu down (503) for ~4 minutes. Not the
rotation -- at memory 512Mi that pod runs for months but CANNOT RESTART:
startup (Prisma + Next.js) peaks over the limit and is OOMKilled (exit 137).
Raised to 1Gi IN THE MANIFEST, not just live, per .claude/rules/manifests.md
("never bare kubectl set -- the next sync reverts it and the fix silently
disappears").
THE GATE: KNOWN_BACKLOG is removed from manifest_bearer_gate.py, as its own
comment instructed. Red-proofed with a decoy: exit 1 with it, exit 0 without.
An exemption kept this visible for three months and changed nothing.
WHAT REMAINS (operator, and it is bigger than what was fixed): the same
password is still the admin password in ~12 other namespaces -- nextcloud,
paperless, bookstack (a DATABASE ROOT password), tandoor, calibre,
adventurelog, gokapi, qbittorrent, servarr, homepage. Rotating Gitea does not
touch them. Also owed: a kisfenyo Gitea token sits in plaintext in the local
homelab-manifests remote URL and was printed to a session transcript during
this investigation, so it should be replaced regardless (R-580's shape).
NOT a finding: homelab-manifests is private (404 anonymously) and does not
contain the password; ArgoCD's repo credential is a separate token and was
untouched by the rotation.
Two new Hungarian pages, live on push: /adatkezeles (privacy notice) and
/feltetelek (what the free closed test is, and is not). Operator rulings of
2026-10-09: no company exists yet, so the operator is named as a PRIVATE
PERSON with no postal address and no phone; publish before the lawyer has
seen it, because the site was collecting data with no notice at all; the full
ASZF, the impresszum and the review wait for the company (R-802, R-809).
- All 18 pages carry the operator, info@felhom.eu and both links in the
footer. Counted, not assumed: 18 pages found = 18 with both links = 18
structurally valid. English footers say the legal texts are Hungarian.
- The contact form stopped claiming something untrue. The old consent said
"az adatokat harmadik felnek nem adjuk ki" while Resend, Cloudflare and
Google carry the message. Both languages replaced; the link opens in a new
tab so a filled form is not lost.
- site_gates.py NO_TWIN gains the two pages ON PURPOSE, with the reason in a
comment: an unreviewed English legal text would be worse than an honest
pointer from the English footer.
- documentation/legal/{adatkezelesi-tajekoztato,feltetelek}-1.0.md are the
text of record, DERIVED from the published HTML so they cannot drift. The
drafts are kept and marked superseded for the closed test.
FOUR LOAD-BEARING CLAIMS WERE MEASURED, not copied from a vendor or a README:
cookies zero, and no local storage - checked in the browser WITH A
POSITIVE CONTROL (a probe cookie WAS visible to the same
method) after the tracker fired; no Set-Cookie on any response
beacon the exact Umami payload: site id, screen, language, title,
url, referrer - no visitor identifier
Cloudflare DNS only: the public A record 37.191.56.193 is not a
Cloudflare address, so site traffic cannot be proxied
fsn1 Falkenstein, Germany (Hetzner's own location list)
Also measured: felhom-ep0-copy-gc.timer is installed and RAN SUCCESSFULLY
(2026-10-09 08:00, exit 0), so "deleted within 30 days" is true today where
on 2026-10-08 it was written but not switched on.
Three retentions are stated as having NO deadline, deliberately and with the
operator's word: website statistics, web server logs and contact messages
have no automatic deletion, and the pages say so instead of promising a date
nothing enforces. Every other period is enforced by configuration and cited.
No placeholder survived onto either page (0 of "[[", control: the draft still
has 36). The impresszum and the full ASZF are NOT published.
R-813 -> NARROWED. R-915 unblocked: the operator enters the two URLs in the
Meta app's Basic settings; the Live switch stays a separate decision.
R-917 and R-920 -> DEFERRED on the operator's (c): park, publish as posts
once posting starts.
Second Facebook task of the day. Page settings changed by hand in the operator's
Chrome; every edit read back from a channel other than the one that made it,
because Meta's toasts have lied here before (2026-10-08: "A modositas nincs
mentve" arrived with a partial save).
- Contact (B1) was ALREADY SET, by the operator, before this run: Graph reads
emails ["info@felhom.eu"] and phone "+36702378499". Verified, not typed.
- Place (B2) already city-only Budapest, as the operator chose. Service area
REFUSED: Facebook offers the field but its picker has no "Magyarorszag",
only cities. Control: "Szeged" returns Szeged. Left unset rather than
narrowed to Budapest, which would shrink a coverage claim nobody authorised.
- Categories (B3) DONE: Informatikai vallalat (kept first, the only one shown)
+ Internetes ceg + Szoftverceg. Facebook's Hungarian list has no IT-support,
IT-consulting or cloud category; eleven terms searched, and the one true
match is a repair counter, which the fences rule out.
- Hours (B4) CANNOT be set and need not be: Facebook requires a street address
first, which the fences forbid. Measured on the rendered page with controls
present -- Zarva 0, Nyitva 0 while Budapest 1, Informatikai vallalat 1. The
Page will never show "Zarva".
- Messenger FAQ (B5) REFUSED -> R-920: the automation does not exist for this
Page. Catalogue holds exactly three templates; search "kerdes" returns none
while the control "uzenet" returns two. COPY.md section 5 is written anyway,
questions verbatim from gyik.html and answers condensed from each question's
own answer, and waits like section 2 does (R-917).
- Link preview correct in both languages, re-scraped once each; only the
expected fb:app_id warning, deliberately not fixed. Both report HTTP 206
where a plain curl gets 200 -- Facebook's scraper, preview complete.
fb_probe.py read now also reports emails, phone, category_list, location,
single_line_address and hours, ONE FIELD PER CALL: a batched fields= list fails
whole when any member is unreadable, which would let one refused field hide the
other five. Refusals are logged verbatim and never retried; "null" and "not
returned" are logged apart. hours is never returned by Graph, which is why B4's
read-back had to come from the rendered page.
Also corrects the "Business & legal" header, which read 12 rows (P2 5) over a
section holding 11 (P2 4); with R-920 it is 12 (P2 4, P3 1, P4 7). Only the
section this commit edits -- the other drifting headers belong to a session
that owns the register.
No post, no invite, no money, no app or portfolio change, website unchanged.
Read phase passes (Page 1360018983863273, CREATE_CONTENT/MODERATE/ANALYZE, page token PAGE expires_at 0, three insights
metrics alive on v26.0). Write test: removal check accepts Meta's code-10 'Object does not exist' (fixed without a row,
4 tests); run 1 evidence kept. R-914 READY, R-915 narrowed to Live mode.
scripts/facebook/fb_probe.py (stdlib, read + write-test, no real-post command) with tests; read run twice:
SYSTEM_USER, expires_at 0, /me/accounts empty, so D/E did not run and nothing was posted. Findings, redacted
evidence, CONTEXT decision home, STATUS item, scripts CHANGELOG, REPORT-facebook-page-api.md.
Host page "Operator Actions" card: run off-site backup now, run a check now
(fixed job list), stop / extend (1-30 days) a deletion countdown. POST
/hosts/{id}/operator-action validates against the CLOSED list before
storing (unknown -> 400, no row), stores operator_actions(id, customer_id,
action, arg, requested_at, requested_by, done_at, outcome, message), logs
who pressed (channel + address) and bumps the box's intent. The report ACK
lists pending rows as operator_actions until the box's
operator_action_results closes them (matched on id AND reporting
customer); each closed row becomes a hub-minted operator_action event
(stored, never dispatched). Unanswered after 24 h: expired. A customer
RESET cancels pending rows. Wire gate: new root + field-by-field mirror
(controller report.OperatorAction) — needs the controller commit first.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
KernelDue / KernelNotify (09-20 h Budapest, one per 20 h, max 3, registered
address, only an accepted mail counts) / os_update.kernel {kver, tonight}
(no mail, no step) / layer kernel ingest + operator events / Approve kernel
set after every ring-0 box booted it healthily after a night stage / two
System page cells. 11 §5.11 written; §5.10 status corrected (proven).
Installer uninstall knows the two GRUB generators (unreleased).
Evidence: audits/kernel-lane-2026-10-07/ (red-proofs, boot timing).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
sysfacts reads the agent's top-level guest_disk_trim stanza (schedule + per-guest
last attempt: vmid, last_attempt_at, ok, bytes_trimmed, mounts, duration_seconds,
last_ok_at, error) into a field-by-field mirror. The System page's new 'Last disk
trim' column shows the last successful trim and the GiB it freed; amber when the
newest attempt failed (error shown) or last_ok_at is older than 14 days (judged on
the success time, never the attempt time); '—' when the agent sends no stanza.
wire_contract_gate: SUBTREE_MIRRORS checks guest_disk_trim field by field BOTH
ways against sysfacts.DiskTrim; decoys (ok renamed, last_ok_at dropped) in
test_gate_decoys.py. Decision 139.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
iso_bootstrap_gate.py runs scripts/iso/test/bootstrap-modes.sh in felhom-iso-assistant:trixie
(staged copy, read-only mount, --network none), registered fast=False in repo_gates.py so the
pre-push hook and CI (both --fast) never run it (decision 147). No docker / no image / docker
error -> exit 2 NOT CHECKED. Every green run is followed by a built-in decoy: the harness must
FAIL a bootstrap whose pairing banner never paints (R-496 shape), or the gate convicts the
instrument as blind. Docker-free decoys in test_iso_bootstrap_gate.py (fake docker on a
one-directory PATH), run from test_gate_decoys.py (COVERS).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
controller/offbox-rename kept its EXEMPT entry after R-425 gave it decoys, so
the debt list read longer than the debt. The gate now prints a STALE EXEMPTION
line for any exempt gate that has a decoy (named, not convicted: the decoy lands
in a sibling repo, and failing this repo's push for it would couple the two).
A check in test_gate_decoys.py plants a stale entry and asserts it is named;
seen to fail with the notice removed.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
DONE/IDEA were searched over the whole state cell, so an open ROADMAP row whose
state named ANOTHER row's verdict (`READY — split out after R-86 CLOSED`)
escaped as shipped — R-87's shape from R-378, one file over. The gate now reads
register_table.leading_verdict(state); all 28 current rows classify unchanged.
Decoys: that shape (seen to PASS against the old gate), an id present in the
register only as prose (seen to PASS when `have` is loosened to any mention),
the existing suffix-id row; genuine: a row with a counterpart, and an `idea` row
asserted BY DESIGN — R-424's hole, accepted by the operator 2026-10-05.
EXEMPT drops `felhom.eu/one-register`.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
Four cases in test_gate_decoys.py, planted inside the real DUE-CHECKS block:
an orphaned check whose id survives only in prose, the due-TODAY boundary with
due-tomorrow as the genuine article (today pinned via FELHOM_GATE_TODAY to a
date before every real check), and the block's marker named in prose (exit 2).
Each convicting decoy was seen to PASS under a mutation of the gate (row match
loosened to the bare id; `<=` to `<`; duplicate-marker refusal removed). The
gate itself is unchanged — no hole found. EXEMPT drops `felhom.eu/due-checks`.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
The R-185 check counted the bare word `felhom-backup-target-apply grant`, so the
dry-run echo stood in for a deleted real grant (2 resolutions, 2 "grants"); now
only path-qualified invocations at a command start count. The age check matched
a commented-out install; now comment lines are excluded. Decoys (plus a version
const in a new hub sub-package, and a comment naming the identifier that must
pass) live in test_gate_decoys.py; all three convicting decoys were seen to PASS
against the pre-fix gate. EXEMPT drops `felhom.eu/hostinstall`.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
admin