Fifth Rule 3 population (function declarations) widened, censused ids stripped, and the gate states its own scan boundary; hot-reload refusal ids stripped

[os-project-manager]

Fixes #13156
Fixes #13179

Family PR executing the twin Class-1 adjudications of 2026-08-29 (director seat, 13:50 + the operative 14:55 pair): A + C on the chain head, strip + C, root extension deferred on the member. One branch, per-member commits, independently verifiable halves. Session: https://claude.ai/code/session_01KX8wnyjStaZcuMyAMNsy3N

Chain head — the fifth population (commits d7e22a3d, 4dd3e5e3, afb72654)

A — the widening. Plain function DECLARATIONS join Rule 3's sink pass in scripts/check-doc-authoring.mjs, under the exact #13002/#13151 narrow clause — a declaration is transparent only when the function itself sits in a recognised customer-facing position, which for a declaration means its NAME is consumed from one (seeded or closed over by collectTextSinkConsts, the same fixed point the hoisted const-arrow spelling rides). Never an unconditional crawl: an unconsumed declaration and one consumed only by an unrecognised call stay unswept, pinned as precision negatives in --self-test.

Bucket choice + rationale (required by the adjudication): the new population gets its own functionDeclared bucket with its own blindness floor, not a fold into functionBuilt. Reasoning is #13151's per-bucket-floor principle applied one step further: the floor exists to catch exactly one clause rotting back to undefined, and the arrow/expression members (217 recognised strings on the real tree) would hold a shared floor up indefinitely while the declaration clause rotted — the precise silence this population was found by. The pre-existing options-factory clause (buildsStrictObjectOptions) is form-agnostic, already reached function declarations before this PR, and keeps functionBuilt; only the name-consumption clause is new. On the real tree the new floor sits at 133 recognised strings.

Reproduce-first, measured: widen first, then strip. At d7e22a3d the widened gate is RED on the real tree: 4 violations — 2 functionDeclared (listPositionFieldReferenceMessage in data/filter.zod.ts:296, normalizedMemberMessage at :1434) and 2 message via the new closure into declarations (SUBMIT_REDIRECT_RULING, FORM_VIEW_FEATURES_RULING in ui/view.zod.ts). After the strip commit the gate is GREEN with every bucket floored (message 1037 · strictObject 3349 · tombstone 771 · describe 8524 · functionBuilt 223 · functionDeclared 133).

Self-test teeth proven by two-leg ablation (mutate, confirm on disk by anchored grep counts, run, restore, prove restoration by HEAD-blob hash match; no build leg exists on this path — the gate runs from source): neutralising the recogniser entry branch fails the declaration REDs and the floor cases while the closure-pin case stays green; neutralising the declaration registration fails all three. Restoration verified: worktree blob equals HEAD blob.

Census re-derivation — a finding, not a blocker. The relayed 47 literals / 51 ids / 14 files (measured on the #13002 branch at c0d50d612) does NOT reproduce on current origin/main: the broad unconditional-crawl census reads 15 literals / 15 ids / 6 files (main moved in the interim — data/field.zod.ts now contributes zero), and the adjudicated narrow instrument reds on 4 of those plus the 2 hoisted consts above. The strip covers the full broad-census population anyway — all 15 in-declaration ids plus the 2 const-borne ids, 17 total across 8 files — because every site is same-audience refusal/warn prose and the ruled remedy is by population, not by count. After the strip the broad census reads 0/0/0.

Strip discipline (#12522 keep-the-ADR-id charter): ADR ids, protocol versions, error codes (INVALID_FILTER / 400) and ruling dates stay; an id that was the whole parenthetical takes the parenthetical with it; genuinely internal load-bearing references moved to adjacent code comments (filter.zod.ts, view.zod.ts). No tombstone had the issue id as its only reference, so nothing needed escalation.

Twins, red-then-green, never weakened: 42 assertions across 7 spec test files red against the stripped tree, 607/607 green after re-pinning — each re-pin a customer-resolvable anchor from the NEW text (ruling dates, ADR anchors, prescription sentences), plus negative id pins at the two ruling-const doors. Two predicate-style pins (.some(w => w.includes(...)) in compose-stacks-key-loss.test.ts) were caught by RUNNING the full mention set after the parse sweep's expect-callee filter missed them — re-pinned the same way (35/35 green). The objectql and service-analytics assertions the id sweep flagged pin messages produced OUTSIDE the scanned root (objectql/src/registry.ts, service-analytics/src/comparand-shape.ts) and are deliberately untouched — they are evidence for the member card's revival census, below.

C — the boundary output (lands once, serves both cards): the gate now prints its own Rule 3 scan boundary on every verdict — the root (packages/spec/src), the explicit sibling-packages-not-scanned statement with the deferral pointer, and the recognised sink-shape list — derived from the same constants the scan reads, pinned derived-vs-derived in --self-test (root from SPEC_SOURCE_ROOT, buckets from the seen map the floor iterates).

Member #13179 — the hot-reload strip (commit 1c749681)

The ruled strip: the plugin-registration refusal's tracker id — the id the #13151-verified twins at hot-reload.test.ts:259,302,310 pin — removed from both registration-door strings (RETIRED_STATE_STRATEGY_GUIDANCE, the distributedConfig entry). Declared bounded same-class extension: the sibling id in the same refusal table's watchPatterns entry and in the startWatching() removal notice is stripped in the same stroke. All four bounded-fix conditions hold: same defect class (same table, same door, same audience), mechanical remedy pinned by the same-day adjudicated charter, same file already on this claim's declared surface, same twin suite with no new verification face. ADR-0049 enforce-or-remove, the spec/core versions and the scheduleReload migration call remain as the customer-resolvable anchors — mirroring the spec-side parse door, which was already id-free with a negative pin.

Twins red-then-green: 8 assertions red (10/18 passing), 18/18 green after; every id pin re-pinned to a content anchor plus negative id pins at all three doors. Repo-wide section-7 sweep for the member's id set: the only remaining test literal is an expect failure-label, not a content pin.

Root extension is NOT in this PR — deferred by the adjudication with a codified revival condition, and not smuggled in via the C output: the boundary line states the limit instead of moving it. The revival condition, however, now measurably FIRES: the diff-derived twin sweep plus a targeted source sweep found same-audience id-bearing refusal/warn/lint prose outside packages/spec/src well past the adjudicated threshold (about eleven candidate sites, e.g. objectql/src/registry.ts:1036, objectql/src/having-filter.ts:186, service-analytics/src/comparand-shape.ts:598, lint/src/validate-react-page-props.ts:434, metadata-protocol/src/protocol.ts:19879). Filed for triage as the instrument card the adjudication prescribed — #13297 (which remains open; it is out of scope here). #13179 is not affected: its own strip stands regardless of how triage grades the wider census.

Verification (all at afb72654, the pushed head)

• Gate union derived by node scripts/pm/dispatch-gates.mjs (no paths; provenance line names this repo at afb72654): 51/53 green, plus check:type-check-debt green after the full 70-package build (30 ledger entries re-measured, none above recorded) and check:nul-bytes green. The two non-green results are the gates' own designed NOT MEASURED refusals, not reds: check-test-completeness.mjs (exit 3 — wants a saved turbo run test log; instructs recording NOT MEASURED locally) and pm/check-half-states.mjs (exit 3 — the container's GitHub token is the proxy placeholder; no reading). CI measures both.
• check:doc-authoring green including --self-test (the edited gate's own suite; no external test file names it — verified by repo grep).
• Tests: spec twin files 607/607; compose files 35/35; core hot-reload.test.ts 18/18; objectql targeted 208/208 (including the untouched out-of-root registry twin); lint targeted 76/76; service-analytics targeted 126/126. @objectstack/spec typecheck green (check:test-typecheck: OK, test layer compiles; debt ledger unchanged).
• Clause-2 posture: content limb no (prose text edits + instrument widening; zero accept/reject change — the widened gate's only behavioural face is its own verdict). PATH limb fires on the spec/src strips, so this PR parks at DRAFT with needs:contract-review; the review chain owns enqueue.

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 2 package(s): @objectstack/core, @objectstack/spec, touching 18 documentable anchor(s).

⛔ 1 release-owned page(s) name something this change touched. These are read-only:

• content/docs/releases/v17.mdx (via stripLegacyApiMethods (symbol))

content/docs/releases/ is RELEASE-OWNED (AGENTS.md "Documentation Guardrails"): release
notes are written centrally at release time, and a code PR that edits them is the exact PR
that guardrail exists to stop. They are still audited — read-only. If one of them is actually
wrong, file an issue or open a dedicated docs-only PR; do not edit it here.

What this run could not see

• 1 anchor(s) matched too much of the corpus to be a work list: sys_user (literal, 29 pages)
• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 132 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 74049254d47bd0edd2a2fcd732dcc01c91504f10 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from fba5b6576265dfa3f8be5d7977d274929aaa9803 — the merge of head afb72654e8e8a76fba283d60fd1c4e1700eb3cfa into base 74049254d47bd0edd2a2fcd732dcc01c91504f10, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin fba5b6576265dfa3f8be5d7977d274929aaa9803 && git checkout fba5b6576265dfa3f8be5d7977d274929aaa9803
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 74049254d47bd0edd2a2fcd732dcc01c91504f10 afb72654e8e8a76fba283d60fd1c4e1700eb3cfa && git checkout -B drift-repro 74049254d47bd0edd2a2fcd732dcc01c91504f10 && git merge --no-ff afb72654e8e8a76fba283d60fd1c4e1700eb3cfa

node scripts/docs-audit/affected-docs.mjs --json 74049254d47bd0edd2a2fcd732dcc01c91504f10

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 74049254d47bd0edd2a2fcd732dcc01c91504f10 → pass the list as
args.docs, on the commit named under Which tree this was computed on.