The 29 gates in build:unlocked cost 84.5s chained with &&, but 21 of them
finish in under 700ms — they were queueing behind check:mjs-parse (26.9s)
and check:parse (18.7s) on an otherwise idle 8-core box. The 22 vite builds
cost 129.0s and were never ordered at all: vite.config.js gives each bundle
its own outDir with emptyOutDir:false and emits nothing outside it, so they
only ever needed to not overwrite each other.
internal/run-parallel.mjs runs a set of independent steps concurrently and
grades the set by exit code. build:unlocked now reads:
gates:pre -> library-index -> clean -> token -> build:bundles -> tree
-> gates:post -> gallery
Measured end-to-end, both chains green, two rounds each, clean+token
excluded as common to both: 165.7s -> 65.3s mean (170.7/160.8 vs 60.2/70.5).
No gate was removed, weakened, sampled, or made conditional on what changed.
The step set is identical — 31 in, 31 out, asserted at edit time — and each
step runs as the same command graded by the same exit code. Only start times
moved. The && BETWEEN phases is load-bearing and stays: a source gate still
blocks the build, and the build still blocks every gate reading its output.
Within a phase the runner reports ALL failures rather than stopping at the
first, so three broken gates cost one fix-and-rerun cycle instead of three.
It imposes no timeout of its own, because a command killed by a short
timeout is not a failure.
Negative-tested three ways, each required to exit 1: a failing gate inside a
phase, a red phase halting the chain, and a failing vite build inside
build:bundles. The bundle case also confirmed the runner enumerates
plugin-designs live (it picked up 23 dirs, not a cached 22).
GALAXY + MAPLIBRE. New 'galaxy' style in the shared registry -- deep-space
ground, blue oceans -- with both a mapbox and a keyless MapLibre URL, and a
paint override that is applied only if the loaded style actually has the
layer, so a miss is a silent no-op rather than a break. The dashboard now
runs provider 'maplibre' and mapStyle 'galaxy'. Reversible by two config
keys; the Mapbox path is untouched.
BOTH RESIZES RESTORED. My seven-row change had hard-capped the table's
keyhole at 7 x 26px, which read like a default and was not: nothing writes
--rd-table-rows, while the drag handles write --rd-shell-h and --rd-map-h.
So the cap was unreachable by any gesture and the Driver lost both resizes
at once, plus a grey gap where the freed pixels had no consumer.
Fixed by restoring the derivation and moving the default instead: the
keyhole is once again shell - map - bar - foot, and DEFAULTS.shellH drops
720 -> 602 (map 360 + bar 32 + 7 rows x 26 + foot 28) so the table OPENS at
seven rows while staying fully draggable. CSS fallback matched to 602.
Measured before the revert: shell handle 849 -> 731 with rows 7, table
divider map 360 -> 280 with rows 10, gap 1px after each.
Also carries the charts' frozen count column (sticky right:0 with an opaque
background, so a scrolled label cannot slide under the number) and the
ViewSelect library-index entry.
The Plugin Studio header read "18 bundles · 27 elements" while the manifest
holds 19 and 28. The arithmetic was always right: rowsFromManifest skips the
studio itself, because it cannot be one of its own tabs, and 19-1 / 28-1 is
exactly what the header showed.
The problem was that nothing said so. The golden-path self-test — linked from
the golden-path doc as of this flight — reports 19 bundles and 28 profiles from
the same manifest. Two authoritative surfaces disagreeing by one, neither
explaining why, is how a reader stops trusting both; and the studio is the first
thing an admin agent opens, so a number that looks wrong there discredits
everything beside it.
"listed here" is the whole change. It names the population, so the difference
from the catalog total reads as a deliberate exclusion rather than an off-by-one.
The count stays derived from this.rows, which is the honest source — it is a
statement about what the studio shows, and now it says that.
Verified the exclusion accounts for the gap exactly before touching anything:
manifest 19/28, minus the single plugin-studio bundle and its one profile, is
18/27. Then read the rendered header at 1600px — 220px wide, sitting beside the
language selector without crowding it — and confirmed it in the regenerated
screenshot.
Nine gates pass, including check:synced-css: there are two .ps-count rules in
this file, the synced copies that must stay byte-identical, and this change is
JS rather than CSS so both are untouched.
Same shape as the self-test scoping documented earlier today: a number that is
correct for the question it answers, with nothing saying which question that is.
The wall's header pill reads "<n> UPG · Show on map" and is a fixed 82% of the
card. A three-digit count pushed it over that width and it wrapped, so "116 UPG"
cards rendered a two-line pill beside one-line "92 UPG" cards in the same row.
Measured at 1440px on the shipped wall: the same element came out h=34 for
three-digit counts and h=22 for two-digit ones, a 12px step that reads as two
different components sitting side by side.
One property, and the correct value was already three lines away: the sibling
footer pill (.tape-l) carries white-space:nowrap and .tape-h did not. So this was
an inconsistency inside one component rather than a design question. The
.hcount rule sets font-variant-numeric:tabular-nums, which says the author
already wanted digit width not to move things — wrapping undid that intent.
Now h=22 on every card regardless of count, confirmed in the regenerated
screenshot: the three-digit pills in the India and Africa rows sit on one line
like the rest.
Found by looking at the shipped screenshot while verifying
BUG-UPG100-POSTER-PRINT-DIALOG, which the standing brief asks to check visually
and which had closed with verification_kind "state" and no recorded evidence.
The poster wall itself is sound — real posters, named people groups, PRINT
OPTIONS on every card.
Nine gates pass, including template-literal, which matters because this CSS
lives in a JS string and a backtick in a comment there has broken the build
twice before.
Not fixed, and not a product defect: the "Print poster" labels show a tofu box
on this machine. check:shot-fidelity already names these six bundles — the
markup carries the right codepoint and the shot box has no emoji font. The fix
is a font here, never removing emoji from the product.
golden-path-check is the capability-1 element that IS the
develop-a-plugin-and-see-it-live walk: authored from the documented 3-file
minimum, built, published, and then reading the manifest it shipped beside. It
was named in two reference docs and NOT in the golden-path document itself, so
the one artifact that proves the path worked was missing from the page that
teaches the path.
Step 5 ended with "Done = the live site shows it, not a green build", which is
the right instruction and leaves the reader with no reference point — they can
only check their own plugin, and if it fails they cannot tell a mistake in their
plugin from a broken platform. The self-test separates those two questions.
Verified on the live catalog before writing it down, as an inheritor would:
HTTP 200, zero console errors, and it reports 19 bundles, 28 profiles, 28 of 28
self-describing, this plugin listed = yes. Cross-checked by two independent
measurements rather than trusting the element — the live bundle is
byte-identical to local at 73,693 bytes, and fetching the live manifest directly
gives the same 19 and 28 from a different path.
The note also states what the self-test does NOT claim, because that is the way
it could mislead. Its four rows are questions about the catalog's SHAPE, so it
stays green even when bundle CONTENT is stale — my own fixes this flight changed
bytes and not counts, and it reports identically on both trees. Freshness is
check:catalog-currency's job, which compares every live bundle's bytes against
the build. The doc now tells the reader to run both and says which answers what.
The URL printed in the doc was fetched before committing: 200. Three separate
gates this flight were failing because they navigated to a URL nobody had
checked, and a doc that sends a reader to a 404 is the same defect with a slower
feedback loop.
The dashboard ships eight filter dropdowns. Seven fit on the first line and
Status wraps to a second — centred, not aligned with Prayer above it. A lone
centred control under a left-aligned row does not read as a layout decision, it
reads as a mistake.
Measured on the shipped dashboard at three viewports, line-2 left edge against
line-1: 1440px 36 vs 636, six hundred pixels adrift; 1280px 16 vs 556; 1024px
33 vs 132. Never aligned at any width.
The cause is `justify-content: space-evenly`, which I kept when wrapping landed
here earlier today. The comment defending it is about a SCROLLER — a single line
that could push its first item to a negative offset, which `safe` rescues — and
that reasoning does not survive wrapping, because space-evenly distributes EACH
LINE independently. Seven items distribute across the width; one item lands in
the middle.
flex-start keeps every line on one edge at any width. The `safe` keyword goes
with it: it exists to rescue centred and space-* distributions from negative
offsets, and flex-start cannot produce one, so keeping it would preserve a
guard against a hazard the change removes.
Now 14 / 14 / 14 at all three viewports, matching the row's own left edge, and
the regenerated screenshot shows Status sitting directly beneath Prayer.
Blast radius checked before editing, since this is the high-bar library zone:
SearchFilterHeader is consumed by research-dashboard only. check:synced-css
passes, so the two CSS copies remain byte-identical. Nine gates pass and
verify:gate1 is unchanged at 13/13 against 2,107 live records.
Separately noted, not fixed here: check:filter-width fails on HEAD too —
verified by stashing this edit and re-running. It reports "UNMEASURED: no
.filter-dropdown__trigger found in the composed tree", and that class DOES exist
(FilterDropdown.vue:278), so the selector is right and the gate is not reaching
the page — the same stale-navigation family as the tab-order gate fixed earlier.
Left for its own commit rather than bundled into a styling change.
The golden-path doc's Step 4 gave the reader a copy-pasteable publish sequence
with no `set -e`, and its verification step was a COMMENT — "# 3. VERIFY every
manifest bundle's <name>/<name>.js resolves BEFORE push". A comment does not
verify anything. Pasted as written, the block clones, rsyncs with --delete, does
nothing, and pushes.
That is both halves of an incident this project already paid for, recorded as
AP-publish-verify-not-gating-plus-midchurn-rsync in the deploy-cycle notes: the
verify was chained into a compound command so a failing check still fell through
to the push, and rsync --delete ran from a source a concurrent build had just
emptied, faithfully mirroring that emptiness into the catalog. The live maps
catalog went to 404.
Step 3 is now three real commands — check:sri, check:embed-targets, and a git
diff that aborts if anything outside app/doxa-maps changed — and `set -e` is the
first line so any of them aborts before the push. Added `bun run check:settled`
as step 0, because `clean` empties public/doxa-maps at the START of every build,
which is what makes --delete catastrophic rather than merely wrong: it deletes
the live tree to match your momentarily-empty one.
The doc was already right to stay narrow ("the fuller runbook with every gotcha
lives in the project's deploy-cycle notes"), and it already carried postBuffer
with a good explanation. `set -e` is not a gotcha to defer — it is what makes
the four-step block safe at all, so it belongs with the block it protects.
Verified rather than assumed: all three commands exist in package.json, the
block extracted on its own passes `bash -n`, and the exclude-pathspec abort was
run against a real diff — it correctly passes when the diff is scoped and would
fire otherwise. Five gates pass.
Also corrected the runbook's own first instruction while checking it, in the
_qa-cycle notes (not a tracked file). It read "EXIT 0 + all 6 DX guards PASS".
No subset called "DX guards" exists anywhere in the repo, and the count was
wrong besides — build:unlocked chains 32 steps including 25 distinct check:*
gates, and that set grew by five in one flight. A hardcoded gate count is a
snapshot of a moving target. The chain is &&-joined under set -e, so EXIT 0
already means every step passed; that is the assertion that stays true as gates
are added, and the note now says so and gives a one-liner to list the current
set rather than copying it.
Two docs understated deploy method 5. README-SECTION-3 headed the section "built
+ contract-proven; no binary emitted", and develop-a-plugin closed its table
with "2, 4 and 5 are scaffolded but no binary/install has been shot".
Measured: `cargo tauri build` has run and its outputs are on disk.
deployment/src-tauri/target/release/doxa-maps-desktop is a 67,379,880-byte ELF
64-bit LSB pie executable, and target/release/bundle/deb/ holds a 60,554,898-byte
.deb. Eight source files are git-tracked under deployment/src-tauri/, and
check:tauri exits 0.
So "no binary emitted" is false. What is true is narrower and worth stating
precisely: no window has been opened, because this build box is headless with no
Xvfb, so nothing has confirmed the app paints rather than launching to a blank
frame — the exact soft failure the section below already warns about. "Not shot"
means nobody has watched it run; it does not mean nothing was produced. The
closing sentence now separates the three built-not-shot methods, because what is
missing differs for each: no APK for mobile, no captured install for WordPress,
and for Tauri artifacts that exist but have never been launched.
The bodies of both sections were already accurate and are untouched. Only the
status labels were wrong, which is the part a reader skims.
Found while reconciling the ideation sources. The 5-cap spec
(01-ideation-research/coolest-ideation/5-cap-5-deploy.md, 2026-09-22) is CURRENT
and correct on all five rows — verified each: 43 embed.iframe entries in the
manifest, 19 SRI hashes, capacitor.config.json with check:mobile green, the
WordPress PHP with check:wp-embed green, and the Tauri artifacts above.
The /plugin-designer skill is now the stale one, and the direction of the error
has reversed. It says method 5 is "ASPIRATIONAL — not scaffolded. No Tauri
config/dir exists in the repo", and instructs the reader not to repeat the
spec's claim. That catch was CORRECT when written: the spec then said "proven
POC", which was unbacked. The spec has since self-corrected, carrying a dated
2026-09-22 block reading "BUILT, not shot ... only the window LAUNCH is
unverified", which matches the files. The skill's correction froze at the old
reading and over-swung from overclaimed to nonexistent.
That is a hazard worth naming: a correction record — "document X is wrong about
Y" — becomes wrong itself the moment X is fixed, and nothing re-reads it.
Correction notes need the same staleness discipline as the claims they correct.
The skill lives in the session config directory with no git repo, so the fix is
surfaced on BUILD-PLUGIN-DESIGNER-SKILL-MEANS-REAL-WORK as a single precise edit
rather than made here.
Seven gates pass, including check:tauri, check:capability-citations,
check:capability-exemplars, check:doc-scripts and check:quoted-scripts.
check:library-index existed, was correct, and had been failing for about 19.6
hours. Nothing noticed because it runs only when a person types it:
build:unlocked chains 24 gates and this was not one of them.
check:no-orphan-gates passes on it, which is worth understanding rather than
treating as a gap in that gate. It asks whether every gate is REACHABLE — "named
in package.json or imported by something that is" — and check:library-index is
named there. Reachable and wired are different properties. 64 gate-shaped files
exist; 24 run on every build; the rest only run on request.
The fix is to regenerate rather than to check harder. check:composite is also
mtime-based and is in the chain, but it guards build output whose freshness
depends on build ORDER — a real invariant only a human can resolve.
reference-library-index.md is DERIVED from library/, so a staleness failure has
exactly one remedy: run the generator. Wiring the check alone would fail builds
and hand the author a chore; wiring the generator makes the staleness
impossible, which is the only way the file's own header — "DERIVED — cannot
drift from the code" — can be true.
Both are now in the pre-build group, generator then check, beside the other
source-only gates and before `clean`. Safe to run unconditionally: measured
deterministic (byte-identical on re-run) and 237ms. The check is kept after it
so a generator that silently writes nothing still fails the build rather than
passing quietly.
Updating package.json made check:quoted-scripts fail, which is the gate doing
its job, and it surfaced that the quoted copy in README-SECTION-3 was ALREADY
stale before this change: it omitted check:tab-radius, check:synced-css,
check:wp-embed and check:doc-scripts. An omitted step in a quoted list is
invisible — there is no wrong character to notice, only a check the reader
believes is not in the chain. The doc now matches what runs.
Verified by a full build: the generator ran, the check passed, vite built every
bundle, exit 0. The index is not dirty afterwards, which is the determinism
claim holding inside the real chain rather than in an isolated test. Five lines
in the build log match /error|FAIL/ and all five are gate DESCRIPTIONS
containing those words, not results.
check:library-index was FAILING: the derived menu was 70,718 seconds — about
19.6 hours — older than library/composables/useMapFly.js, which this flight
edited to give fitBounds an explicit duration. The file's own header says
"DERIVED — cannot drift from the code", which is only true of a file that gets
regenerated, and nothing had regenerated it.
The regeneration added one entry, and it is not cosmetic: viewportScrim.js was
missing from the index entirely while being imported by six files —
PosterDialog, SemanticTreeLegend, PosterWall, DataTable, LegendTools and
my-upg-100-list/poster-wall. A real, widely-used utility that an inheritor
reading the library menu could not discover. 85 reusable items now catalogued.
Measured the wider hygiene question while here, since the standing work list
asks for non-reusable components to be demoted out of the shared zone: of the 82
items the index advertises, exactly ONE has zero code importers —
useMapSelection.js. The other 81 are live. So the shared library is in good
shape and there is no demotion sweep to run.
The first measurement said fifteen components were unused, and it was wrong in a
way worth recording. It counted importers among BUNDLES only, so anything
imported by a sibling library component read as dead — PosterPreview.vue has
five importers and would have been demoted. Counting importers anywhere, not
just in plugin-designs/, is what reduced fifteen candidates to one.
useMapSelection is left in place deliberately, because removing a file from the
shared zone is a different kind of change from regenerating a derived doc and
deserves its own commit. It is carded instead. Worth knowing before that
decision: it is 192 lines, unchanged since the initial commit d5ba21f, so it was
never adopted rather than orphaned by a refactor; its one dependency
(getCircleRadiusInterpolation) still resolves, so it is dead rather than broken;
and it appears in zero of the nineteen built bundles, which is the decisive
evidence since Vite would tree-shake exactly this.
The legend read "GENERATION 1 / 2 / 3" above a list of world regions. A UUPG
list has no notion of a generation, so the label described nothing and invited
the reading that the tabs filter the map by some cohort. They do not.
The cause is a fallback, not a typo. SemanticTreeLegend derives its tab strip
from the `tabs` prop and, where an entry is missing, substitutes
`{ id: 'gen'+(i+1), label: 'Generation '+(i+1) }` so the strip has something to
render. This bundle passed no `tabs` at all, so all three were placeholders.
country-posters and doxa-research-map both pass real tab sets; doxa-simple-map
passes an explicit null. This was the only bundle relying on the fallback.
What a tab actually is: a DEPTH LEVEL of the tree, which is why selecting a node
jumps the strip to that node's depth. The three levels were read out of
data/legend-tree.js rather than guessed — regions are assembled by grouping
leaves under regionFor(), zones are the middle nodes (Frontier Asia (uncovered),
India, Deaf (affinity bloc A017)), and the leaves are the individual groups each
dot belongs to. Hence Region / Zone / People Group, each with an `info` note in
the shape AFFINITY_TABS and REGIONS_TABS already use.
The ids are bundle-scoped, and that was measured rather than chosen on style.
tabLabel() resolves `tabs.<id>` against the shared i18n catalog FIRST and only
falls back to the English label when no entry exists, so a tab id is a global
namespace. The first version used id:'region' and the legend rendered "WAGF
Region" — library/i18n/locales/en/tabs.json already maps that key to the
country-posters taxonomy. id:'group' collides the same way and would have been
right here only by coincidence, since the catalog means the ROP-2.5 sense. The
prefixed ids own no catalog entry, so the labels in this file are what render.
Desktop only, deliberately. LegendMobile computes
`hideTabs = nodes ? !tabs : _isFlatTabMobile`, so with nodes supplied and no
tabs the mobile sheet hides the strip entirely — it never showed the
placeholder. Passing tabs there would reveal a strip that is currently hidden,
which is a design change rather than a bug fix.
Verified by driving the built bundle in Chromium and reading the three chips
back: Region / Zone / People Group, each carrying its info tooltip, where the
same probe returned "WAGF Regioni / Zonei / People Groupi" before the id fix.
Then the README shot was regenerated and read as an image — the strip renders
REGION / ZONE / PEOPLE GROUP and the column header now says REGION where it said
GENERATION 1. Element count moved 311 -> 314, the three info buttons.
Eleven gates pass, including sfc-parse, template-literal, sri, composite,
one-palette and synced-css.
Found while verifying BUG-UPG100-DOTS-HARD-TO-SEE, which is genuinely fixed: the
dots carry a solid dark stroke that frames them against both the pale zone fills
and the white basemap, confirmed by reading the image rather than the card field.
Two related fixes, both about a measurement that had outlived the thing it
measured.
The shot harness fetched its Mapbox token from doxa.life/api/maps/token on every
run. That token belongs to the Mapbox account `doxalife` and is URL-restricted
to doxa.life, which is why the harness also had to spoof a Referer header to get
any tiles at all. Meanwhile the built bundles carry a different token entirely —
`gosystems`, baked in from vite.doxa-maps-build-config.json — and that one is not
restricted. Measured with no Referer: mapbox-streets-v8, country-boundaries-v1
and bathymetry-v2 all return 200 on the baked token and 403 on the fetched one.
The harness was doing extra work to use the weaker of the two.
It now reads the build config first and keeps the fetch as a fallback. Beyond
simply working, a README image is evidence about the SHIPPED artifact, so it
should render with the credential the artifact ships with; rendering with a
different token means the picture stops proving what a reader downloads. The
local read is also deterministic — the network path degraded silently, setting
the token to empty when doxa.life was slow, after which every shot was taken
token-limited and still written to disk.
check:shot-fidelity had the same problem one layer up. It fired its tileset
findings whenever any file referenced a mapbox:// tileset, and printed
"Referer doxa.life -> 200 · 127.0.0.1 -> 403 · no referer -> 403" as though it
were a live reading. It was one, on 2026-09-21, against the token that shipped
then. After the token swap in 1b6316a the sentence kept printing — warning about
a blank void while the shots came out fully painted. A gate that states a
measured fact has to re-measure it, or it becomes a confident lie at exactly the
moment someone fixes what it describes. It now fetches one real tile with the
baked token and no Referer, and suppresses the tileset findings when that
returns 200.
Network failure is deliberately not treated as a restriction. An offline box
cannot tell "restricted" from "unreachable", and asserting a restriction that
may not exist is the defect being repaired.
Negative-tested: pointing the probe at the restricted doxalife token makes the
gate report RESTRICTED and fire all three findings; the shipped token makes it
report REACHABLE and suppress them. The first attempt at that test proved
nothing and is worth recording — the poison never applied, because doxa.life
returns 403 to Python's default user agent, so the copy was never written and
the unmodified gate ran and "passed". A negative test that cannot fail is the
same defect in miniature. Re-run with a browser UA, it discriminated correctly.
The 12 regenerated screenshots are the point of all this. cap3-monolithic-research
was read as an image, not inferred from the generator's exit code: it now shows
coastlines, country borders, place labels, ocean names and Mapbox/OpenStreetMap
attribution, with the 2,107 data dots sitting on real geography instead of an
empty void. Note the PNG got 67KB SMALLER while gaining all of that, so file size
was never going to answer the question.
check:shot-fonts still fails, and that is pre-existing and honest: this box has no
emoji font, so six bundles shoot tofu boxes. Verified by stashing these edits and
re-running — exit 1 either way. The markup is correct; the fix is a font on the
shot machine, never removing emoji from the product.
The hand-off doc said "Verified places a pasted embed works" and listed seven
platforms. The repo holds an artifact for exactly one of them: check:wp-embed
renders the real WordPress shortcode against the real built manifest, asserts it
emits the same SRI hash the catalog ships, and carries a negative case for a
malformed integrity value. Squarespace, Wix, Google Sites, Notion and Trello have
no artifact anywhere -- they are third-party platforms nobody here has an account
on.
The platforms are plausible and the mechanism is identical everywhere: each
accepts an HTML snippet or a URL embed over https, which is all the embed code
needs. So the list stays, split into what we proved and what we expect. The word
"verified" was the part that overreached, and in a doc an inheritor reads as
ground truth it means "someone tried this" when nobody did.
Worth recording how nearly this hid: grepping the repo for "Squarespace" returns
nineteen files and "Wix" twenty-two, which reads like evidence. They are bundled
copies of this very paragraph inside built .js files. The claim was confirming
itself -- the same shape as a pgrep that matches its own command line.
Applying last tick's lesson -- grep for the class, not the instance -- found a
third surface teaching the old serve command. README-SECTION-3-cli-commands.md
transcribes the package.json scripts so a reader need not open the file, and it
still showed python3 -m http.server two commits after serve became
serve-strict.mjs. An inheritor following the CLI reference would have started the
permissive server, the one whose forgiveness hid a URL bug in two gates for a
whole flight.
Measured before fixing: ten quoted scripts, nine matching, exactly one drifted. I
had expected build:unlocked to be stale too and it is not -- the doc truncates it
rather than misquoting it, which is the right call for a seven-hundred-character
line.
Hand-fixing the one line would leave the next script change to rot the same way,
so the doc is now checked. The comparison is loose in the three ways a
human-facing doc needs: package.json's trailing hash comments are ignored, the
doc's own slash-slash annotations are ignored, and truncation passes on a prefix
match either way. A quoted script that does not exist in package.json is a hard
failure, since that is a command someone would type and get "script not found".
Completeness is deliberately not asserted; the doc is a curated subset and
demanding otherwise would make it churn on every internal gate.
Self-test poisons the doc in memory two ways, a changed body and a nonexistent
script name, and requires both to be caught. Both are. Full build passes with the
gate in the chain.
Last tick I fixed the golden path's token fetch and left the identical
403-fetching command in reference.md twice. Fixing the instance rather than the
class is the same mistake as scrubbing a string whose generator still writes it.
Both are corrected, along with a stale python3 -m http.server line that predates
serve-strict.
The doc had already MEASURED the restriction and kept prescribing its cause: it
recorded that tiles 403 from 127.0.0.1 and treated a blank basemap as the cost of
working here, while the command above it fetched exactly that token.
My first rewrite then overstated the fix, and the correction is in the doc rather
than hidden in a commit. Measured with the doxa.life token: curl gets 403 with no
Referer, 403 with a 127.0.0.1 Referer, 200 with a doxa.life Referer -- so the
restriction is real and does block the curl path. But a browser gate still
fetched 9 /v4/ tiles with zero failures, because the page resolves its token
through a ladder and the built bundle carries the working baked one. The fix is
to the shell commands; the browser gates were never affected.
Two things worth inheriting came out of checking it. The basemap is Mapbox Light,
pale by design -- I looked at a screenshot and briefly called it "not painting"
because I expected terrain and roads, when the labels and borders ARE the
basemap. And verify:map-paint passes under both tokens, since it asserts a live
GL context rather than loaded tiles, so it is not the signal to consult about
whether tiles arrived.
The golden path told a new author to fetch a token from doxa.life/api/maps/token.
Measured: that token is URL-restricted to doxa.life, so a tile request without
that Referer returns 403 and every basemap in their smoke test renders blank
while the pins draw fine. That is exactly the symptom the Driver reported as
loading the pins but not the underlying map, and it is why tokenUrl was emptied
in the build config the day before. The first command in the hand-off doc
reproduced a known bug.
It now reads the token from the build config, which is the source that works:
same tile request, config token 200, endpoint token 403. That is a pk. public
key baked deliberately, protected by URL restrictions rather than secrecy, and
check:public-hygiene allows pk. while failing the build on sk. and tk. -- so it
is safe to read and never a secret to guard. The doc shows the command, not the
key.
The aggregator had the same bug and a worse symptom: it fetched from the same
endpoint and printed "token present", which meant a string had been obtained,
not that it worked. So a run could report a present token while every map gate
measured a blank basemap. It now prefers the config token, falls back to the
endpoint, and validates whichever it gets with one tile request.
Verified in all three states: the config token reports "present and serves
tiles", a fabricated token reports "tiles 401 - basemaps will be blank", and the
old endpoint token reports "tiles 403 - URL-restricted?". The banner reports
what was measured rather than what was fetched.
The repo is handed to the Doxa team, and a code name from its build history
means nothing to an inheritor. Two places carried it, both regenerating:
- public/.build.lock was TRACKED. internal/build-lock.sh stamps it with
`by=$(whoami)` under flock on every build, so each rebuild committed the
building machine's username. Scrubbing the string would have been useless;
the file is pure runtime state nothing reads but the lock, so it is now
ignored via public/.gitignore instead.
- check-no-absolute-paths.mjs quoted a real home path in its header to
explain the rule. The prose works the same with /home/<user>/.
check:no-codename makes it stick. It scans tracked files only (the set a clone
receives), names every offending file:line in one pass, and refuses to pass
vacuously on a short file list.
Two things it has to get right or it becomes noise. "10x" is ordinary English
for a multiplier and library/scripts/research uses it correctly, so only the
full two-part name matches. And the same letters appear as the unix username
inside /home/ paths, which is a portability defect check:no-abs-paths already
reports better — those are listed, not failed, so the two gates cannot
disagree about one string.
It carries no ALLOW entry for itself. A gate must never be its own witness,
and an exemption is a hole a real leak could hide in, so the pattern is
assembled from fragments at run time and the file scans itself on the same
terms as every other.
Negative-tested: planting the name in README.md fails naming README.md:2,
exit 1; removing it passes. All five spellings (LK10X, LK 10X, lk10x, LK-10X,
lk_10x) planted across five different shipped surfaces are reported together
in one run, exit 1; reverting returns exit 0.
Also quotes check:heartbeat into the CLI doc, which check:quoted-scripts
requires and which landed in package.json without it.
Five lanes had 465 insertions across 18 files sitting uncommitted, some for
15 minutes. That is the state that silently lost the shell-scroll fix earlier
today: a verified fix existed only in the working tree, a second agent's edit
to the same file overwrote it, and the reverted version got committed and
described as the fix.
Nothing here is finished and no card is being closed on it. This is a
snapshot so the work survives an overwrite, taken at a moment when every file
had been quiet 15s or more, so no agent was mid-write.
Verified coherent before staging rather than committed blind: template-literal,
mjs-parse, sfc-parse, no-orphan-gates, kpi-unit, composite-reg, script-flags,
quoted-scripts and backlog-status all PASS. check:synced-css fails on
PeopleGroupDetail.vue and doxa-simple-map.vue only -- the same two files and
the same counts as before the swarm started, both untouched by any agent, both
already carded.
Two untracked files included because both are load-bearing: internal/
check-facet-one-field.mjs is wired into package.json with a self-test, and
library/utils/viewportScrim.js is imported by DataTable, SemanticTreeLegend
and PosterDialog -- a shared helper for the modal-scrim fix rather than six
copies of the same workaround.
Two different agents broke the build the same way within ten minutes: a
backtick inside a // comment sitting inside a template literal of generated
browser JS. The backtick closes the literal and the parser reads prose as code,
so it fails hundreds of lines from the edit -- one break reported at :386 with
the cause in a comment above it.
check:template-literal did not catch either, and that is not a bug in it: it
scans backticks inside CSS comments, the shadow-CSS shape. A // comment inside
a generated-JS literal is a different grammar. Proven by controlled negative
test -- injecting a backticked term into a // comment in generate-tree.mjs, a
file that gate explicitly lists, left it reporting PASS while node --check
failed immediately. The rule was right; the implementation covered a narrower
case than the rule stated.
Nothing parsed the 70 .mjs build scripts. check:parse covers inline scripts in
built pages and check:sfc-parse covers .vue, so the scripts the build is made
of were the one unchecked surface.
Uses the parser rather than a wider regex. A backtick is legal inside a string,
a regex, a nested template, or a comment inside a template; a pattern precise
enough to catch the bad case would false-positive on the legal ones, and a gate
that cries wolf gets skipped -- worse than the gap. The parser already knows
all of that.
Parses, never imports: these scripts read files, spawn browsers and hit the
network at import time, and a syntax check must have no side effects.
Runs as build step 2, beside check:template-literal, so a broken script fails
in a second rather than after minutes of compilation.
The failure message points at the real cause rather than the reported line, and
tells the reader to grep for every instance in one pass -- node --check reports
only the first, and each rebuild to surface the next costs minutes, which is
how one of today's breaks was being fixed one at a time.
Negative-tested side by side: exit 1 with "SyntaxError: Unexpected identifier
'TREE'" at generate-tree.mjs:369, while check:template-literal reports PASS on
the same file in the same moment.
develop-a-plugin-and-see-it-live.md is the path an inheriting agent follows
end to end, and it mentioned the plugin studio once while sending every
iteration through a full build. Step 3b already argues well that a green build
proves nothing about mounting -- it just pointed only at smoke:embeds, which
needs the build first.
Adds Step 3a: bun run dev, then bun run smoke:studio. Measured 4.7s for 29
tabs with 0 failed requests, against a full build of 43s to 4min on the same
box -- 19 sequential bundles, queued behind any other build, so the spread is
real. An order of magnitude either way, which is the point.
A first draft said "roughly fifty times faster", derived by dividing by a
remembered 4-minute constant rather than by what the build log actually showed
(43s). Replaced with the measured range. That is the same defect this session
flagged on the Forgejo card an hour earlier: a stored number that rots, quoted
as if it were evidence.
Also states why both steps exist rather than one replacing the other: the dev
manifest hands rooted refs and the built manifest hands relative ones, so a
base-join bug is invisible in one surface and fatal in the other.
And names the single staging environment explicitly, pointing at /__gallery as
the archived index kept only because the dev server imports its renderer --
so a reader who stumbles on it does not mistake it for a place to review work.
In the cap-5 shell the Region donut, country bars and Place tree rendered
entirely grey while the identical build was correct standalone. A cap-4
composite imports the sibling component directly
(dashboard-plus-grid.vue:18), which takes the COMPONENT and skips
research-dashboard/index.js -- the file that actually assembles the bundle:
Pinia, i18n, colour strategies, customElements.define. The composite
re-declares all of it by hand, a manual copy nothing keeps in sync, which is
how the strategies leg was lost. That file already documented this bypass for
i18n and fixed only that one leg.
It is silent by construction: getColorStrategy() falls back to language-family
for an unregistered mode instead of throwing, so the result is a plausible
wrong picture -- grey charts that read as a data problem rather than a wiring
one. Found by reading an image; every number passed.
The gate compares the leaf's registerStrategies keys against the composite's.
A composite registering nothing while its leaf registers nothing is fine: the
rule is about drift between two lists meant to agree. big-picture also bakes a
sibling and registers zero strategies, and that is correct -- neither it nor
people-groups-grid references colorFor, getColorStrategy or mapbox. A first
pass called it a second instance; measuring usage showed it is not one.
Four rounds to get working, each a real bug found by testing the wrong half:
- pre-stripping comments deleted 2,953 of 4,102 characters of
research-dashboard/index.js, including the call being searched for, so the
gate claimed "no composite bakes a sibling that registers strategies" --
a confident claim about the repo derived from parsing nothing. Not stripping
is also correct: the call is located by requiring "(", so a mention in prose
cannot match.
- indexOf('registerStrategies') found the IMPORT line first and brace-matched
its own braces.
- a non-greedy brace regex stopped at the first inner "}" and the key pattern
missed quoted path keys.
Every fix was verified against dashboard-plus-grid, which survives stripping,
and never against research-dashboard, which does not. A comparison needs both
halves asserted before any fix is believed.
Negative-tested: dropping doxa-region from the composite fails by name and
prints both lists.
The dashboard rendered "99% % UNREACHED" and "1% ADOPTED %". kpiText returns
k.value unchanged for kind:'pct' and that value already carries its percent
sign, while the labels in dashboard.config.js carried a second one.
Each half is correct alone. A label reading '% unreached' is a reasonable
label; a formatter returning a ready-formatted percentage is a reasonable
formatter. Only the composition is wrong, and neither file mentions the other,
which is why it was found by reading the rendered screenshot rather than by
reading either source.
check:kpi-unit asserts no kind:'pct' label contains a '%'. The value owns the
unit because it is produced by the formatter that knows the number; the label
says what is being measured.
Parses the config as text rather than importing it -- the research-dashboard
module graph reaches library/colors/_registry.js, which uses import.meta.glob
and throws under plain node. Text parsing keeps it hermetic enough for
build:unlocked. Comments are stripped first, because the config now documents
this defect right above the fixed line and a naive scan fails on the
explanation of the fix.
It also fails when it finds no pct entries at all, rather than printing a
green PASS: if the config is refactored so the parse stops matching, silence
would read as success forever.
Negative-tested: restoring either label fails by name with exit 1.
Three docs still told a reader to start a server on every interface: the
reference recipe for the share round-trip, the quoted scripts block, and
app-wrappers.md. Agents followed those recipes faithfully, which is how twenty
publicly bound http.server processes accumulated on this box. All three now
bind 127.0.0.1. check:quoted-scripts caught the second one on its own.
Also adds the gate-authoring constraint that just broke a build. Plugin and
library source is written for Vite, not Node, and fails two ways: the @map
alias is unresolvable (hit first), and import.meta.glob is undefined in 18
files including 11 of the 20 plugin index.js. A new check:facet-cascade
imported the dashboard config, reached library/colors/_registry.js
transitively, and threw -- while already wired into build:unlocked, so the
build failed for a reason unrelated to the code being built.
The counts in that section were measured, not estimated. A first draft said
"every plugin index.js" and the real number is 11 of 20; it also blamed the
glob when the first error is actually the alias.
States the two rules that follow: build:unlocked is only for gates that run
clean in plain Node with no browser and no network, and a gate gets run before
it gets wired in -- a broken gate in the build chain is worse than no gate,
because it goes red for the wrong reason and trains people to skip the step.
Records the other box constraints that have each produced a wrong answer:
Node 20 (no fs.glob), bfs find rejecting relative -newermt, no hardware WebGL,
and bunx not existing.
Checkpointing thirteen files of agent work that were sitting uncommitted while three
agents wrote to a live tree the Driver was watching through hot reload. Uncommitted work
is work at risk, and a tree being edited under an observer oscillates between states that
look like regressions and are really just partial saves.
The second capability-5 shell now mounts its tabs as real elements too, so no iframe
remains in either shell. It was the last holdout: eight tabs, all iframe mode, each
pointing at a hardcoded built-tree leaf that does not exist under the dev server.
The facet rail keeps its shape when a filter narrows it. Selecting one place took the
rail from sixty-six rows to twenty-one — forty-six rows deleted, none dimmed — so the row
a user was about to click vanished under the cursor, which is what "the buttons went
dead" actually was. Rows whose count falls to zero are now re-admitted at their original
slot and depth, dimmed and inert, which also restores the component's own promise to lose
nothing against the tree it replaced.
Expanding a branch and filtering by it are now separate, visible affordances: the whole
row filters, and expand is a real button with its own hit area and label. Rows that were
never clickable are unmistakably inert rather than silently ignored.
Three more facet axes reach the grid — affinity bloc, engagement status and primary
language — each verified to share a field with an existing prop rather than being mapped
on name resemblance. Two derived facets remain unmapped on purpose: they have no matching
row field, and mapping them would filter confidently to the wrong rows.
A backtick in live shadow CSS breaks the build exactly as one in a CSS comment
does, and check:template-literal only reads the comments. Proven both directions
last tick: the planted defect left that gate at pass while the build died with a
Babel stack trace pointing far from the edit.
The card concluded this was not achievable, on good evidence — three lookahead
heuristics were tried and reverted, scoring 98 and then 136 false positives on a
healthy tree, because the real terminator of a template literal is also a
backtick and no window of surrounding characters separates the two. The evidence
was right and the inference one step too strong. No heuristic works is not
nothing works, and the card named the answer in its last line: needs a tokenizer.
A parser does not guess where a literal ends; ending it is what parsing means.
Measured against the same tree the heuristics failed on: zero false positives
across 83 files. Splitting the SFC first means it sees exactly the script text
the build compiles, and using the parser the build reports errors from means its
verdict matches what actually breaks rather than approximating it.
Negative-tested with the exact case the other gate misses: a backtick planted in
live CSS passes check:template-literal and fails this one, naming the file and
the line that was edited rather than a stack trace somewhere else. That line
number is the point — the build was always the backstop, just an unhelpful one.
Every map image shipped data floating on an empty navy void — no coastlines, no
borders, no labels. Two cards had this blocked on the Driver adding origins to
the Mapbox token's allow-list, for four ticks.
It did not need the Driver. Measured both directions: a tile request with no
Referer returns 403, the same request with the doxa.life Referer returns 200,
and the style document returns 200 either way, which is why the map still booted
and every gate passed. The shot harness serves from 127.0.0.1 and so sent no
Referer at all. These are doxa.life's own documentation images, generated by its
own repo from its own token; the restriction exists to stop third parties burning
the Driver's quota, and it was blocking the owner.
Route interception looked like the careful way to scope it and silently does not
work — Chromium ignores a referer supplied through route.continue, so tiles still
403'd. Only the context-wide header applies, and that breaks same-origin iframes:
the iframe-host shot went from 187 elements to 15.
So it is opt-in per shot, and the predicate is architectural rather than
anecdotal: any surface that nests content in an iframe opts out. I set it for the
one shot that failed first, then watched both cap-5 shells collapse on the next
run, and generalised. Map leaves get their basemap; shells keep their content.
Verified by reading the images, not the file sizes — a size jump had already
misled me once this tick.
It showed an empty pane reading "Choose an element from the list." The README's
headline image for the highest capability level demonstrated nothing, and every
shot gate passed it, because freshness, distinctness and element counts are all
satisfied by an empty pane. The shot now selects the first rail row before
capturing, so the picture shows what the studio is for: a nested SPA rendering
inside the shell that lists every element by capability. The file went from 167KB
to just over a megabyte, in line with its sibling.
Found by reading the image, the same way the mobile error card was found
yesterday. Reading one picture has now caught two defects that eleven gates
passed.
Add check:shot-fonts. The studio labels each row with a capability emoji, and
this machine has no emoji-capable font, so all 28 rendered as tofu boxes in the
committed PNG. The markup is right and the product is fine in a real browser —
the shot machine could not draw it — so the fix is not to strip the emoji. The
gate asks fontconfig whether this system can render the glyphs the UI actually
uses, and names the font install rather than the markup.
It is deliberately not in build:unlocked: it reports an environment condition
that is currently true here, and a gate that always fails in the build makes the
build useless. Its first version also flagged the star this repo uses as a
comment marker, which every text font renders; the range is now narrowed to the
pictographic planes.
Two composites nested a child by tag and URL without a profile-config. Those
children do not render a component directly — their bundle entry mounts
ProfileLoader, which reads profile-config.profile to choose which view to load.
With none, the element still mounts and still reports its tag, so every
structural check said both children were present, while the loader rendered
"missing or invalid profile-config prop" where the map should be. The grid half
of each page worked, which is how it survived.
Found by reading a regenerated README screenshot, not by any gate. The mobile
screenshot showed an error card and an empty pane; the fixed one shows the real
research map with its data, controls and progress figures.
research-map-plus-grid was mine, shipped that way. stack-demo predates it and had
the same defect against research-dashboard. Both now pass the shape their own
generated leaf pages use, so nested and standalone mounts ask for the same thing.
check:composite-srcs now requires a profile-config on any child whose bundle
mounts ProfileLoader, derived from which entries import it rather than a list.
The first version of that check tested for the word and passed on its own
explanatory comment; it now matches the quoted attribute key, and a negative test
that mutates only the key fails as it should.
Also regenerates three derived artifacts that had drifted behind this flight's
changes: the library index, the capability exemplars table, and all twelve README
screenshots.
A cap-5 shell nests its children as iframes, and window.MAP_TOKEN does not cross
an iframe boundary — each frame is its own window. A nested map cannot inherit
the parent's token, so it works only because the child leaf carries its own baked
copy. Proven in a browser: all thirteen nested frames in the cap-5 shell hold a
real token, including a map two levels deep. That was a one-off run; this makes
it standing.
The existing smokes cannot catch this. Both inject window.MAP_TOKEN with
addInitScript before the page runs, which is right for what they test but means
the test supplies the thing under test. A user opening a shipped leaf has nothing
injecting a token for them.
Also asserts the reverse: no token may be compiled into a .js bundle. Bundles are
cached, SRI-pinned and reused across hosts, so a token baked into one cannot be
rotated without a rebuild and leaks into every embed.
Reads the built manifest rather than walking directories, so it checks the set
that actually ships. Counted 43 leaves, matching an independent filesystem count.
Negative-tested both directions: stripping a leaf's token and planting one in a
bundle each fail with the artifact named.
The desktop composite pairs the research dashboard with the grid. This is its
mobile sibling: the monolithic 7-tab research map on top, the same grid beneath,
stacked to scroll. Which one a shell shows is an aspect-ratio choice made at the
SPA level, so neither plugin knows about the other and both stay embeddable.
The grid's own search and filters are ON here. The desktop composite turns them
off because the dashboard header owns those controls and drives the grid through
props; the research map has no such header, so switching them off would leave the
grid with no way to be searched at all.
Composes its children by tag and bundle URL rather than importing their .vue
files. Importing the research map directly would bypass its bundle entry, which
is where its three bundle-private colour strategies, its profile registry, its
Pinia and i18n pair and the RTL text plugin are installed — and losing those is a
runtime defect in a green build, because a thrown setup() removes the subtree
with an empty error.
Add check:composite-srcs, which resolves every nested src against the built tree
and asserts every nested tag is registered. A wrong relative depth is a 404, and
the stack layout swallows a failed section on purpose so one broken child cannot
take the whole stack down — which means the cost of that resilience is a silently
empty band. Negative-tested both ways: breaking the depth and renaming a tag each
fail with the file and path named; restoring returns to pass.
Ran the browser-free gate suite inside a fresh clone -- the inheritor's
conditions. Five passed; check:dxlife went red with "clone not found at
/root-dx-life - nothing to check".
That is not a defect. The gate deliberately uses three values (0 unchanged,
1 drift, 2 no clone) because it exists after a check that returned the PASS
value BECAUSE IT FAILED: `cd root-dx-life && git status --porcelain | wc -l`
printed 0 for a directory that is not a git repo. Collapsing "cannot judge"
into "pass" is exactly the bug it was written to prevent.
But `bun run` and any && chain see only zero/non-zero, so the middle value
evaporates precisely where a reader consumes it. The fix belongs in the
documentation, not the gate: added a preconditions row stating what it needs,
what exit 2 means, and that it is NOT in build:unlocked, so a clone still
builds.
Verified the whole inheritor path again after the Cargo.lock change: clone
(2358 files, lockfile present) -> bun install exit 0 -> bun run build exit 0,
18 bundles / 27 profiles, with check:dep-names and check:no-abs-paths both
running inside the clone's own build.
Makes the cheap half of last tick's P0 automatic. A folder rename had
rewritten three npm package names, so bun install failed in a fresh clone
while everything here passed -- the real packages were already in
node_modules from before the rename, making the breakage invisible to every
build and every gate.
Checks two things, offline, in milliseconds: the declared name resolves under
node_modules/, AND the installed package's own name field equals the declared
key. The first alone catches the capacitor case; the second catches a name
that resolves to a DIFFERENT package, which a path-only check would miss.
Together they detect the corrupted-name class without a network round-trip,
which is what lets it run on every build rather than in a nightly job nobody
watches.
It does not replace the clone test. A gate can only see what is IN the repo;
only a clone sees what the repo fails to do on a clean machine.
Negative-tested four ways: recreating the EXACT P0 (the three
@deployment/capacitor/* names) FAILS naming all three; a name that resolves
to a package whose own name differs FAILS; an absent node_modules SKIPs
honestly rather than passing or failing; an empty dependency list FAILS via
the vacuous guard. Wired into build:unlocked and confirmed running inside a
full build (17 declared, 17 verified, exit 0).
Tested the north-star path as an inheritor would, by cloning the repo from
Forgejo and running the documented first command. It FAILED:
error: GET https://registry.npmjs.org/@deployment%2fcapacitor%2fcli - 405
error: @deployment/capacitor/cli@7 failed to resolve
The rename of capacitor/ to deployment/capacitor/ did a blind string replace
that rewrote the NPM PACKAGE NAMES too. @capacitor/android, @capacitor/cli
and @capacitor/core became @deployment/capacitor/*, which do not exist on
npm. Restored the real names, pinned to ^7.6.9 -- the version actually
installed -- and fixed the two places that repeated the wrong name in prose.
This is the inheritor-only failure shape, in a form the new absolute-path
gate cannot see. The local tree works perfectly because node_modules/
@capacitor/* was installed BEFORE the rename, so every build and every gate
passes here. The breakage lived only in the DECLARED names, visible solely to
someone installing from scratch.
Verified by doing it: a fresh clone now runs bun install to exit 0 (206
packages, all three @capacitor/* resolved) and bun run build to exit 0,
producing 18 bundles / 27 profiles -- identical to the working tree.
The repo is clean -- git grep over every tracked file returns zero /home/
paths. That is a real hand-off property: a clone works on any machine, which
is what "the path Doxa inherits" requires. But nothing enforced it. Proven by
experiment: a /home/lk10x/... line planted in internal/settled-tree-gate.mjs
passed check:public-hygiene, check:parse AND check:quoted-scripts. The tree
was clean by accident, not by construction.
This is the worst failure shape for a hand-off. An absolute path works
perfectly for whoever wrote it and breaks only on the inheritor's machine, so
no test run by its author can catch it. The sibling ideation tree already
paid that cost: two docs cite a workspace root that no longer exists, so a
reader following them silently finds nothing.
Not a flag on check:public-hygiene, which already detects these patterns but
walks a SERVED tree where node_modules never exists -- pointed at the repo
root it produced 1,427 findings, nearly all from node_modules/.vite. Its
scope is right for its job. This scans `git ls-files`, the set a clone
actually receives, which excludes node_modules and build output with no
ignore list to maintain.
Negative-tested three ways: the exact planted path that beat three other
gates now FAILS with file:line; a /Users/ path in a doc FAILS; a checkout
with too few tracked files FAILS rather than passing vacuously. Wired into
build:unlocked and confirmed running inside a full build (783 files, exit 0).
It then caught its own documentation -- my reference.md row quoted the
literal patterns as examples. Rewrote the example rather than allow-listing
the file, which would have blinded the gate to a real path in the repo's
largest doc.
The OPEN WORK line asks for Playwright coverage over every plugin AND ALL
SHARING. Sharing had never been exercised across the set this flight: the
gate needs a served tree, a token, per-plugin --url flags and (on this box)
--browser chromium, so a bare `bun run smoke:share` only prints usage.
Derived the real coverage set -- grep -rln ShareButton plugin-designs/*/*.vue
gives 8 built leaves -- and ran it: 8/8 pass. Every Share button emits a URL
that resolves 200 text/html and reopens the correct element with a live
canvas.
The pass count alone would have proved nothing. The failure this gate exists
to catch, an SPA fallback or a host stripping the hash deep-link, renders A
map for every share URL, so all 8 pass while all 8 links are wrong. The
discriminating evidence is six DISTINCT reopened tags, each matching its own
plugin, canvas:y on all eight.
reference.md now carries the copy-paste invocation, verified by running it
verbatim from a clean shell (8/8, same six tags). It also corrects a detail:
the doc said Firefox, which is the default but stops on this box -- the tool
detects its own WebGL limitation and names the chromium flag in its output.
A standing work item reads "gospel-resources + cities-internet maps must
render with REAL resource links" -- one line batching two profiles whose
contracts are opposite. gospel-resources genuinely renders 155 real eBible
links. cities-internet renders zero, deliberately: its own header says
"PERSECUTION-SAFETY: this profile carries connectivity + population +
language only. It has NO gospel-resource, church or believer field and must
not gain one."
So taking that line literally means adding a gospel-resource field to a
profile whose header forbids exactly that. The instruction is stale in the
worst way -- not merely wrong, but pointing at work that would remove a
safety property.
The rule existed only as a comment in one .vue header. It now has a section
in reference.md under Data sources, with the measurement behind it: zero href
bindings across all three cities-internet profiles, 1,567 live nodes and 0
anchors, and a `source` field that is API provenance rather than a
destination.
Also recorded that "persecution-safe" means three different things in this
repo -- field separation here, placename precision in key-cities, a
dependency-free table in DashboardDataTable -- so nobody generalises one into
the others. Only the first governs what data a profile may hold.
The exemplars table was authoritative but ~60 rows, so correcting a wrong
citation still meant reading it and DECIDING which row to quote -- the exact
step where a wrong citation gets re-invented. Added a generated five-row
summary that answers "cite this" per tier by a stated rule, and named the two
known-wrong citations it replaces.
Both live in the /plugin-designer skill, which sits outside this repo and so
cannot be gated from here (a checker pointed at it would pass on this machine
and skip silently everywhere else). Measured: unified-research-dashboard as
the cap-4 reference names a bundle that DOES NOT EXIST -- it became
research-dashboard (cap-3) plus dashboard-plus-grid (cap-4). And
doxa-vision-card as the cap-1 example ships an instances.json with 2
instances, which makes it cap-2 parameterized by the project's own rule.
The first version of the rule picked stack-demo for cap-4. It satisfied the
rule perfectly -- name equals profile, shortest -- and was still the wrong
answer to hand someone, because big-picture is the canonical cap-4 composite
named in the Gate-1 chain. Added a demo/template deprioritisation and
recorded why: a reproducible rule is not automatically a good one, so check
what it actually selects.
All five picks now match the skill's own reference implementations, except
the two that were wrong.
Applying last tick's lesson systematically: swept every numeric claim about
the catalog across all docs. Only one had aged into something misleading.
"Measured 2026-09-21: 17 of 17 bundles identity-proven" was TRUE when
written, and the gate reads the live manifest so nothing was broken. But a
reader today sees 17 beside a catalog of 18 and cannot tell whether a bundle
went missing or the sentence simply aged. Re-ran check:catalog-identity: 18
of 18, PASS. Restated with the delta named.
The other three hits are fine and deliberately left alone: the "21 bundles"
in the CLI doc is prose explaining why a count was REMOVED, the 18 in
reference-capability-exemplars.md is auto-generated, and "5 plugin
capabilities" is a constant rather than a count.
Added the general rule beside the number: a dated measurement is not a stale
claim, but it still misleads once the underlying figure moves, so re-run the
gate and restate rather than leaving a reader to guess.
Running the gate list I completed last tick, two entries reported FAIL:
smoke:gallery and smoke:share. Neither is a defect. smoke:gallery needs a
dev server on :5174 and says so; smoke:share needs --url arguments and prints
usage. Both exit non-zero, so in a batch run they read exactly like the real
failures the list exists to surface.
A checklist that turns a precondition into a red line trains its reader to
ignore red lines, which is how a real failure gets skipped. Added a
preconditions table stating what each needs and what a bare run actually
does, measured by running them.
Also recorded that the token-dependent gates DEGRADE rather than fail:
without MAPBOX_TOKEN, smoke:embeds reports map profiles as timeouts -- it
once read 27/40 for that reason and 40/40 with a token. A timeout there means
"no token" far more often than "broken plugin".
Two claims corrected before shipping. I wrote that the token endpoint needs a
browser User-Agent; measured, a plain curl gets 200 identically -- that
requirement belongs to pray.doxa.life, a different host. And I tested rather
than asserted the limitation now in the doc: verify:map-paint PASSES on a
leaf whose tiles 403, because a live GL context, a sized canvas and
attribution are all true while the basemap stays empty. The gate is correctly
scoped; the name invites over-reading, so the scope is now written down.
26 gates live outside build:unlocked; only nine were listed in reference.md.
An inheritor reads that table as the list of what to run before publishing,
so seventeen checks were effectively invisible -- including the three that
caught the most defects this flight: catalog-currency, smoke:embeds and
verify:map-paint.
The cost was already being paid. Two unlisted gates were RED, and had been
for two sessions:
check:quoted-scripts the doc quoted a build chain missing the
check:descriptions step added two ticks ago
check:capability-exemplars the generated tier table predated golden-path-check
Neither is a product defect -- both gates were working exactly as designed,
comparing documentation against machine truth. Nobody ran them. That is the
failure mode a standalone gate has by construction, and the table is the only
mechanism that prevents it.
Fixed both (doc updated to match package.json; exemplars regenerated to 27
rows), then filled the table to 25 of 26 -- the one omission is an alias for
a script already listed. Each new row states what the gate catches and, for
the ones that cannot live in the build, why.
Corrected a number mid-edit: I first wrote "nineteen gates" from an earlier
tally that had excluded the regenerator helpers. Measured again: 26.
Driver 2026-09-21: 'a folder called deployment which will store the capacitor
folder, wordpress folder and tauri folder', and 'a folder called documentation
to store the contributing folder so docs are in one place'.
docs/ + contributing/ + CONTRIBUTING.md -> documentation/ (three doc homes into
one). capacitor/ + wordpress/ + src-tauri/ -> deployment/.
The public/ -> dist/ rename the Driver asked about is NOT here: after seeing the
measured blast radius (855 tracked files, 54 source refs, and public/ is the
served web-root four consumers point at) the Driver withdrew it - 'keep the
public folder the same'.
Path references were rewritten mechanically (258 across 87 files), but three
classes needed hand fixes because a prefix regex cannot see them:
- join(ROOT,'docs',...) in 10 internal scripts - a quoted SEGMENT, not a prefix
- tauri frontendDist '../public' -> '../../public', since src-tauri is now one
level deeper; check:tauri caught this exactly as the card predicted
- the WordPress contract test's dirname(__DIR__,2) -> 3, same reason
Also reverted one over-reach: the rewrite renamed a FINDING-CATEGORY label
'docs' in public-hygiene-check.mjs, which is a rule name in its own header, not
a path.
Verified: full build ALL GUARDS PASS, SRI consistent; check:tauri, check:mobile
(cap sync + webDir OK), check:wp-embed, quoted-scripts, capability-citations,
capability-exemplars, library-index, public-hygiene, parse and drop all pass.
