Pouya LajevardiandClaude Opus 5
0f7595b602
Pouya's five rulings of 2026-08-31, after 64bce10. One commit, because AGENTS.md
entry (aj) covers all five and the review fixes interleave across them —
recorded here rather than left to be inferred.
1. THE EYEBROW, 13 -> 14px, at the one token definition. `--text-eyebrow` is
deliberately NOT an alias of `--text-sm` despite sharing its value: the two move
for different reasons, and aliasing would mean a change to body-meta type
silently moving every eyebrow. Sweep by rendering, 22 pages x 2 widths, 801 mono
instances: 441 carry `.eyebrow`, 420 now at 14px, 21 held at 11px. Zero of 801
below 4.5:1.
Header check, which Pouya asked for explicitly: nothing header-side consumes the
token, `--header-h` is 81px, and the sticky header measures exactly 81.00px at
every width from 66rem up with the CTA on `.header-inner`'s content edge. Nothing
overflowed, so nothing was improvised.
2. /contact/'s LABELS RAISED, AND THE FLOOR CLAIM REWRITTEN BECAUSE THE RULING'S
PREMISE WAS FALSE. The ruling asked docs/02 to record the tagline as "the ONLY
sub-14px text on the site". It is not: the `--text-xs` 12px rung has ten
declaration sites and 130 elements rendering at 12px. So docs/02 records the
claim that IS verified — `--text-2xs` has one consumer in the repository — and
enumerates the 12px rung as a separate treatment. Writing a false claim into the
spec because the conclusion was handed down is the move the rules forbid.
A third block moved that the ruling did not name: /bio/'s ten <h2>s and its strap
were copies of five of `.eyebrow`'s six declarations at 11px — the same escape the
footer headings were. docs/02 justified holding them by the one-page PDF
constraint; measured, `@media print` sets both to 7pt, so the screen size never
reached the PDF and that was never the reason.
Consolidating them DID change the printed sheet, because the missing sixth
declaration was `font-weight`: 500 grew the PDF 89,496 -> 91,151 bytes. Print
therefore freezes 400, the committed PDF stays byte-identical (10 differing
bytes, all /CreationDate and /ModDate), and unifying is Pouya's call.
3. THE REFLOW DEFECT WAS SITE-WIDE AND /bio/ WAS NOT THE BINDING CAUSE.
`white-space: nowrap` on `SiteHeader .brand-name` held all 22 pages 63px over at
320px/root 32. Eight cause-specific fixes, no `overflow-x` anywhere; all eight
produce byte-identical geometry at root 16. `Pill` is marked in docs/02 as the
backstop it is — the real cause was `PracticeCard`'s rem-based padding, and with
it clamped "Construction" goes from 94x220 in six two-character lines to 158x85 in
two.
Three instrument findings now in docs/02, each of which hid a real defect:
`break-word` does not reduce min-content and `anywhere` does;
`getBoundingClientRect()` reports border boxes, so an element sweep cannot see
text spilling outside its own box; and `mobile: true` emulation expands the
layout viewport, so `scrollWidth - innerWidth` reads 0 while the reader still
scrolls sideways.
4. TTL STAMPED `ENABLED` in §7, with `DISABLED` at first verification recorded
rather than overwritten. Q60 narrowed to its second half and OPEN: no record has
been watched to vanish, and `ENABLED` proves the setting, not the behaviour.
R19's sweep found three stale copies outside §7; all now defer to it. R19 itself
was not edited — it points at §7 rather than carrying state, which is the
property that made it work.
5. §4 GAINS THE STRUCK-UNIVERSAL ROW, citing the committed ontario.ca extract and
cross-referenced to `check:claims`'s `struck-universal-q39`. It bars the claim in
BOTH directions: the commercial half is Pouya's attributed position, not a
verified fact. The row immediately caught two places asserting it flatly —
`SiteHeader.astro` and §9 Q33 — both now attributed. `check:claims` unmodified;
still frozen.
REVIEW: adversarial-reviewer, two rounds, 16 findings, ALL ACCEPTED, NONE
DECLINED. claims-auditor did not run (D20). Eight of round 2's ten were defects
in round 1's own fixes.
Round 1's blocking finding was a defect in my own record: I wrote that every
reflow residual was zero at "root 16 and root 32, 286 measurements". Two ways of
simulating 200% text are NOT equivalent, because media-query `rem` resolves
against the DEFAULT font size, not the root element's. Under the method docs/02
itself prescribes the site is 944px over on 21 pages and 508px on / at 1280, and
304px at 1920 — nav clipped mid-word, Practice/Fees/Contact and the CTA
off-screen, WCAG 1.4.4 with loss of functionality. My 286 excluded exactly the
widths where the defect lives.
NOT FIXED, DELIBERATELY: the cause is `flex-wrap: nowrap` on `.nav-list` above
66rem, a locked step-1 decision, and the standing instruction is to stop and
report rather than improvise a header change. It is now a blocking item on
docs/06's cutover checklist, ticked only by ruling on it — "not by re-measuring
it with the method that reports zero".
Round 2 also caught: a 63,743-byte figure that was the `cmp -l` differing-byte
count rather than the 1,655-byte size delta; a lost-navigation list naming
Med-Arb, which is not in the masthead at any width, while omitting Contact; the
tagline's justification left recorded at 13px, where one clause of it is false at
14px (20px of document overflow at 1216 with a seventh nav item); my own label
raise reintroducing a 38px element overflow at the width just ruled on; two
copies of the eyebrow treatment left at weight 400; three stale residual tables;
and the min-content explanation duplicated six times in src/.
VERIFIED, exit statuses read directly, never through a pipe: build 0 (22 pages),
check 0 (0 errors/0 warnings/0 hints), check:claims 0, og:proof 0, check:intake 0,
lint 0, minifier tripwire clean, TODO in dist 0 with a source sanity check.
Overflow: 374 measurements over 22 pages, every one zero, row counts asserted
before reading, with positive controls (119px at width 200, 319px at root 64).
Lighthouse 0, run three times with identical category scores: perf 99 on / and
100 elsewhere, a11y 100, best practices 100, SEO 100 on all indexable pages,
CLS 0.000. / sits at LCP 2.03s against a 2.0s budget, unchanged by this work.
The zsh `$VAR` word-split trap fired twice more and both times read as a clean
pass; caught only by asserting row counts first. Two of my own instruments were
wrong before they were right: a `grep -F '0.875rem'` state check that could never
match because Lightning CSS writes `.875rem`, and a probe whose inline
`!important` was outranked by a running CSS transition.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Md3GndFqWPzK78xAoebsg5
adr.smlcompany.ca
The dispute resolution practice of Pouya Lajevardi — Toronto.
A static site built with Astro, built for deployment to Amazon S3 behind CloudFront by Gitea Actions — see Deployment; the pipeline is not yet proven.
Quick start
nvm use # Node 22 LTS — the floor is in package.json engines
npm install
npm run dev # http://localhost:4321
Scripts
| Command | Does |
|---|---|
npm run dev |
Development server with hot reload |
npm run build |
Static build to ./dist — 22 pages |
npm run preview |
Serve the built site locally |
npm run check |
astro check — type and template errors |
npm run check:claims |
AGENTS.md §4 Forbidden, enforced on dist/. Runs on every deploy |
npm run check:intake |
The intake form's field table against the Lambda's — two on purpose |
npm run og:proof |
Every og:image resolves; every card headline is its page's <h1> |
npm run lint |
ESLint + Prettier check |
npm run format |
Prettier — rewrite files in place |
npm run lighthouse |
The performance budget, all four categories. Local only |
npm run bio:pdf |
Re-renders the committed one-page PDF from /bio/. Local only |
npm run deploy |
Build and deploy from this machine — see Deployment |
Two of those cannot run in CI, and that is stated rather than left to be
discovered. lighthouse and bio:pdf drive an installed browser; the Gitea
runner has none. They are keyboard gates plus blocking items on docs/06's
cutover checklist, and they are deliberately not wired into npm run build
or either deploy path — a check described as running where it cannot is the
defect AGENTS.md Q22 turned out to be.
og:proof and check:intake exist because two facts here are deliberately
duplicated, and a duplicated fact needs a mechanism rather than a comment.
Text baked into an OG card cannot be grepped by check:claims, so og:proof
comparing each card's headline to its page's <h1> is the only thing keeping
card copy inside the claim register. And the Lambda validates against its own
field table, because a server that validates against a list the client shipped
it is not validating.
(npm run check:claims was missing from this table before 2026-08-31, along
with the four added that day. A table of the project's controls that omits a
control is the shape those controls exist to catch.)
Before you contribute
Read AGENTS.md first, and maintain it as you work — it is the living
record of what this project is, what was decided, and why. Then read
CLAUDE.md for the working rules, and the specs in docs/.
The single hardest rule: no factual claim about Pouya, his credentials, his
experience, or his practice ships unless it appears in the verified register in
AGENTS.md §4. This is a public marketing surface, and the site it replaces
contained fabricated credentials.
How work is done here
Pouya decides; Claude Code implements and then adversarially reviews its own
work. Run /build <task> for any substantive change — it plans, implements,
runs two independent review agents on the diff (the claims audit wherever copy
changed), resolves the findings, verifies the build, and records the session in
AGENTS.md. /review runs the review pass alone; /wrap closes a session.
Full protocol and prompt guidance: docs/08-execution-protocol.md.
Deployment
Today, deploys run locally: npm run deploy (scripts/deploy-local.sh).
It runs the same guard, the same three sync passes with the same cache headers,
and the same invalidation as the CI workflow — at this scale the pipeline
changes only how a deploy is triggered, not what it does. The script and
.gitea/workflows/deploy.yml are one artefact in two places: change one, change
both.
.gitea/workflows/deploy.yml is the CI pipeline — Gitea Actions, not GitHub
Actions. It has never run, for two reasons that are not oversights:
- Actions are not enabled and no runner is registered. The Gitea instance is jointly administered, so both need its second administrator (Q23).
(It previously listed a second reason — that the scoped IAM user did not exist.
It exists: adr-sml-deploy, created 2026-08-26, Q22 closed 2026-08-28. See
AGENTS.md §7 for the inventory and the least-privilege verification.)
Its first step is a guard: the run aborts, naming what is missing, if any required variable or either AWS secret is empty. Only emptiness is tested and no value is echoed.
The GitHub Actions equivalent, which uses OIDC role assumption, is kept as
docs/reference/github-actions-oidc.yml.example in case the project ever moves
to a forge that supports it. It sits outside .github/workflows/ on purpose:
Gitea falls back to that directory when .gitea/workflows is absent, so a
workflow file left there with a push trigger would be only conditionally
inert. As an .example under docs/ it cannot be picked up at all.
The pipeline is designed around a long-lived AWS credential, and that
credential now exists. Gitea is not an AWS OIDC provider, so there is no role
to assume: deploys authenticate as a scoped IAM user, adr-sml-deploy, with its
access key in the repository's Gitea Actions secrets. The user was created
2026-08-26 and verified least-privilege by execution — AGENTS.md §7, Q22 closed
2026-08-28. Its access key has never been used (LastUsed null), because
deploys still run locally. (This paragraph asserted the user "has not been
created" until 2026-08-28.) In the meantime the local script
refuses to run as user/pouya, the broadly-permissioned personal user —
see AGENTS.md §10. Two things are meant to bound the risk, and neither is in
place yet:
- The policy must stay narrow. Four actions:
s3:ListBucketon one bucket,s3:PutObjectands3:DeleteObjecton that bucket's contents, andcloudfront:CreateInvalidationon one distribution. NoAction: "*", noResource: "*", nothing outside that one bucket and that one distribution. The AWS account is shared with unrelated projects, including a bucket whose name indicates another business's production database backups — that narrowness is what keeps a compromised runner away from it, and it is load-bearing rather than hygiene. SeeAGENTS.md§10. If a deploy step needs a permission the policy lacks, question the step; do not widen the policy. - The key must be rotated quarterly. It now has an owner and a date: first
rotation due 2026-11-26 (key created 2026-08-26) —
AGENTS.md§12 R17. Create a second access key, update the Gitea secrets, confirm a deploy succeeds, then delete the old one — rotation that leaves the old key active is not rotation, and deleting before verifying leaves no way to authenticate the fix. OIDC would have removed the obligation entirely; it is unavailable, so this is a standing calendar task. (It read "nobody owns that yet" until 2026-08-28; R17 is the owner, and the date is the whole point of the row.)
Full procedure, IAM policy, runner setup, and cutover checklist:
docs/06-deployment.md.