tooling(pm): the claim-side readers — a bulleted Branch: governs, and a Refs-only landing is not H8's full delivery

[claude]

Fixes #16036
Fixes #16170
Fixes #16037

One file, scripts/pm/check-half-states.mjs. No new rows; the family markers are untouched.

#16170 — a bulleted Branch: line governs again

claimedBranches now admits an optional markdown list marker before the directive:

^\s*(?:[-*+]\s+)?>?\s*Branch(?:es)?\s*:

Measured live on #15511, whose claim comment is the malignant MIXED shape: an
undecorated Claim: line (so CLAIM_COMMENT_MARKER, which has no list marker
either, matched and the comment IS a claim) with the branch written as
- Branch: ... under it (so this reader returned nothing and the claim GOVERNED
nothing). The two anchors failed at different points on one comment, and nothing
in the artefact looked wrong.

The claim MARKER is deliberately NOT widened with it. When BOTH lines carry a
bullet the marker misses too, the comment is not a claim to any reader in this
file, and "no comment on the card's thread is a claim comment" is then the TRUE
reading — so only the mixed shape was malignant, and admitting the marker here is
what removes it. Widening Claim: itself is the 2026-08-11 ruling's territory
(不放宽谓词), not this reader's; the self-test pins the marker still false on a
bulleted Claim:.

The refusal message is NOT edited here, and the reason is reported rather than
acted on. The sentence the card quotes — "no comment on the card's thread is a
claim comment" — lives in scripts/pm/check-clause2-carriers.mjs (the absent
limb of cardDeclaration), which this flight does not touch by ruling. Reading
that limb: it ALREADY distinguishes the two states the card asks for, and the
predicate that separates them is CLAIM_COMMENT_MARKER, not the branch parse —
missing says the claim comment is there and the declaration line is what is
owed, absent says no claim comment exists. On the measured #15511 shape
claimRows is non-empty, so that thread lands on missing, whose sentence is
accurate. See the report comment on #16170 for the full reading and the open
question it leaves.

#16036 — a Refs-only landing is not H8's full delivery

H8's MERGED side now reads a second, narrower predicate, prFullyDeliversCard:

export function prFullyDeliversCard(pr, n) {
  if (!prDeliversCard(pr, n)) return false;
  return refsOnlyLinksFor([pr], n).length === 0;
}

refsOnlyLinksFor is H49's own predicate, CALLED rather than restated, so what
H49 owns is exactly what H8 excludes and the two cannot drift into a gap or an
overlap.

The shared prDeliversCard relation is untouched, deliberately: H8's OPEN side,
H31's carrier comparison, claimDelivery and the pairing check-clause2-carriers
derives all want the wide "is there a PR on this card" reading, because a Refs
half still open is live work. Narrowing the shared relation would make that half
invisible to the rows that exist to see it — the #10468 harm, run backwards. The
open-side branch-fallback pin is unchanged and still green.

Pins, all three from the dispatch ruling: a Refs-only merged PR on a card-named
branch leaves H8 clean; a closing-keyword body fires as before; a body-silent
merged PR on a card-named branch fires as before (the fallback's original #10757
population). Refs #N BESIDE a delivery does not make it partial, and a Refs
for another card narrows nothing about this one — bound per card number, as H7
binds.

#16037 — three line-number citations deleted

The header inventory's H1 entry (SKILL.md line 109, SAME write (line 465)) and
selfTest's RELEASE_RULE_LINE comment (SKILL.md line 465, VERBATIM). Numbers
out, quoted text kept, in the same style PR #16040 used for the two ruled
docblocks. git grep for line 109 / line 465 / line 493 over this file now
returns nothing.

Verification

All exit codes captured before any pipe.

Self-test — node scripts/pm/check-half-states.mjs --self-test, exit 0:
2535 cases before, 2553 after. Case-name diff: 20 added, 2 removed, both
forced by the ruling that flips H8's verdict on one fixture and each with a named
successor on the same input:

removed	successor
H49 adjacency: on a head NAMED for the card H8 fires through its branch fallback	H49 adjacency: on a head NAMED for the card H8 is silent too, since #16036
H49 adjacency: …and this row fires beside it — two readings, no double silence	H49 adjacency: …and this row fires there — one reading now, and no silence

Two names survive on a moved subject and are named here rather than left to be
found: H8 branch: a Refs #N-only body delivers via its branch name now asserts
the unchanged prDeliversCard relation (still true of that input), and
H8 branch: …and the finding names the delivering PR now asserts on the
body-silent specimen, which is the case that still produces a finding.

Ablations — on a scratch copy outside the worktree, byte-pinned to the HEAD
blob before each run and restored by blob-hash comparison after it. Each mutation
was confirmed on disk by removed-text and injected-text occurrence counts plus a
blob-hash change, before the run was read:

• revert the claimedBranches list marker: blob 62db8ef to ed7c565, removed-text
  count 1 to 0, injected-text 0 to 1 — self-test exit 1, 6 named failures, all
  the check-half-states: a Branch line written as a markdown bullet parses to ZERO claimed branches, so the claim silently stops governing while the refusal message blames a missing claim comment #16170 cases including the mixed-shape governing one.
• revert H8's merged side to the wide relation: blob 62db8ef to 5778d9e, same
  on-disk proof — self-test exit 1, 2 named failures, the two check-half-states: H8's branch-name fallback reads a merged Refs #N (item k) PR on a card-named branch as a full delivery, so it prescribes dropping pm:dispatched on a lawfully re-dispatched remainder #16036 cases.
• both restored to blob 62db8efec42e3d68e9c7e76ec2c880bf0b1524d9, equal to
  HEAD:scripts/pm/check-half-states.mjs.

The first ablation initially died on a TypeError naming neither row nor mutation;
the second commit on this branch adds the ?? [] row wrapper H34's own note
prescribes, so the reverse run reports NAMED cases. That is why the ablation
numbers above are from the re-run.

Gates — node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack
derived 33 commands at commit c493f98f22; all 33 run, all exit 0;
--ran reconciles: 33 derived famil(ies) accounted for — 33 run, 0 NOT-MEASURED,
exit 0. node scripts/pm/check-governed-merges.mjs --test scripts/pm/check-half-states.mjs
exit 0 — NOT governed — ordinary queue landing applies.

The 8 artifact-roster families whose roster sits under scripts/ or scripts/pm/
(where the derivation's silent verdict is evidence in neither direction) were
run too: 6 green; check:published-readme-exports and check:dts-closure both
exit 3 = PREREQUISITE NOT MET (no package has a dist/ in this worktree),
which is NOT MEASURED rather than red — this diff touches no package source, so
neither can reach it, and CI builds the closure and runs both.

Lint — whole-repo pnpm lint (eslint . --no-inline-config) through
bash scripts/pm/os-verify-lock.sh -c with slot issue-16036:
VERDICT command-exit 0 · held the lock 73s · waited 1s.

Live sweeps, read-only, both boards, before and after. Two scratch trees, one
pinned to the merge-base blob and one to the HEAD blob, both verified by
git hash-object before running.

board	before	after	rows removed by this diff
objectstack-ai/objectstack	299	301	0
objectstack-ai/objectui	173	182	0

Every difference is board drift over the minutes between the paired runs, and
each one is accounted for rather than waved at:

• objectstack: PRs test(service-cluster-redis): pin the WATCH/MULTI abort-retry loop and the versioned-delete MULTI/DEL branch #16054 and feat(objectql): publish SYSTEM_WRITE_ORGANIZATION_REQUIRED_CODE and isSystemWriteOrganizationRequiredError so consumers stop re-spelling the literal #16156 MERGED between the runs. Their H46 rows
  (an OPEN-PR population) left, and their cards [finding] service-cluster-redis's KV contract suite never exercises the WATCH/MULTI abort-retry path nor the versioned-delete MULTI branch — measured by instrumenting the shipped test #15983 and objectql: publish a recognizer for SystemWriteOrganizationRequiredError so consumers stop re-spelling ERR_SYSTEM_WRITE_ORGANIZATION_REQUIRED — instanceof is measured unsound across the CJS/ESM split #14936 gained H8 rows
  naming those exact PRs. Both of those new H8 rows come from closing-keyword
  bodies, so the narrowing correctly leaves them firing — the ruling's second pin,
  observed live. The other movements (H19, H31, H35, H36, H44) are in families
  that read neither claimedBranches nor H8's predicate.
• objectui: H20 on test(spec-tooling): restore the liveness author-lint's array fan-out assertion (#7079) #7689 and [finding] $icontains is absent from analytics TEXT_PATTERN_OPERATORS, so the #5234 comparand fence never covered it on the where door — one operator, two answers inside one package #7693 fired at "~62 min after the claim was posted
  (threshold 60 min)" — a clock crossing, not a reading change — and their H8 rows
  arrived with PR [finding] uninstallPackage unregisters the namespace BEFORE the verb that can refuse — a rejected uninstall leaves the package half-mutated #7970's merge in the same window.
• H8's row set is a strict superset after, on both boards, which is the
  direction the change permits: narrowing delivering can only remove rows.
• H49 reported 0 rows with complete coverage on both boards (63/63 then 64/64
  on objectstack, 13/13 on objectui), which is the reason nothing was silenced:
  no card on either live board currently carries a merged Refs-only landing
  inside H8's merged window. The guard is prospective, and the self-test is where
  it is measured.

One open question, left to the seat rather than decided here

The ruled anchor puts the list marker BEFORE the optional blockquote, so
- > Branch: reads and > - Branch: — the natural markdown for a bulleted line
INSIDE a blockquote, and arguably what "blockquote + bullet" names — does not.
Implemented exactly as ruled, with the residual pinned as a measured fact and
flagged in the report comment. Say the word and it is a one-character move.

skip-changeset: scripts/pm/** publishes nothing from any package.

🤖 Generated with Claude Code

Generated by Claude Code

---

Generated by Claude Code

[claude]

ACCEPT — in-seat review of PR #16202, head c493f98f22 (skills seat, 2026-09-06T14:2xZ; the compare-worktree reading was first taken at 05:4xZ, and re-taken on a fresh worktree at 14:1xZ after the seat's container was reclaimed).

Implemented-by: claude/issue-16036-claim-readers (os-dev subagent, opus under the measured quota exemption recorded on the claim)
Reviewed-by: session_019RfFHiRCSs3JXLK4cwcfox (skills seat)

Verified in a detached compare worktree at the PR head, not from the report:

• One file, scripts/pm/check-half-states.mjs (+185, −32), no new rows, family markers untouched. check-half-states: a Branch line written as a markdown bullet parses to ZERO claimed branches, so the claim silently stops governing while the refusal message blames a missing claim comment #16170: claimedBranches reads with the ruled anchor verbatim — ^\s*(?:[-*+]\s+)?>?\s*Branch(?:es)?\s*: — so the mixed shape measured live on platform-objects(i18n): the zh-CN dashboard helpText leaf goes stale the moment #14586 lands — a fresh instance of the class #14931 names, with every gate green #15511 (undecorated Claim: over a bulleted Branch:) governs again, and the claim marker is deliberately NOT widened: a bulleted Claim: is no claim to any reader in the file, which is the accurate reading and the 2026-08-11 ruling's territory. check-half-states: H8's branch-name fallback reads a merged Refs #N (item k) PR on a card-named branch as a full delivery, so it prescribes dropping pm:dispatched on a lawfully re-dispatched remainder #16036: H8's MERGED side reads the new prFullyDeliversCard, which CALLS H49's own refsOnlyLinksFor rather than restating it, so what H49 owns is exactly what H8 excludes; the shared prDeliversCard relation is untouched, and the reason given (H8's open side, H31 and claimDelivery need the wide reading, else a live Refs half goes invisible) is the right one. [finding] check-half-states.mjs still cites SKILL.md by line number in three places (header line 109 / line 465, selfTest line 465) after the two ruled docblocks dropped theirs #16037: the three SKILL.md line-number citations are gone — git grep for line 109 / line 465 / line 493 over the file returns 0, quoted text kept.
• Gates at the head: check-half-states --self-test exit 0, 2553 cases (2535 before; 20 added, 2 removed, both removals forced by the ruling flipping H8's verdict on one fixture and each with a named successor); check-governed-merges --test scripts/pm/check-half-states.mjs exit 0 (NOT governed ⇒ queue landing). PR checks on c493f98f22: 24 success, 11 skipped, 0 failed.
• Report hygiene: dev report on check-half-states: H8's branch-name fallback reads a merged Refs #N (item k) PR on a card-named branch as a full delivery, so it prescribes dropping pm:dispatched on a lawfully re-dispatched remainder #16036 with the literal os-dev-report line, JSON closes; PR body opens with the three Fixes lines; no model identifier; skip-changeset read back; one MCP call declared (the check-half-states: H22's closed-card pass is TRUNCATED at the 12-page quota ceiling on this board, so the closed-residue census is silently short #16217 filing).

Rulings on the two open questions (the dev implemented as ruled and asked, which is the correct order):

1. Anchor order — A. The anchor stays as landed. The only measured live specimen is the bare bullet; neither mixed order exists on either board today, so admitting > - Branch: (B) would be a widening for a shape nobody has hit. The residual pin the dev left, naming the ruling as its authority, is the legible form of that gap. If a seat ever writes > - Branch:, that reading is a new card, not a rider.
2. check-half-states: a Branch line written as a markdown bullet parses to ZERO claimed branches, so the claim silently stops governing while the refusal message blames a missing claim comment #16170 part (b) — A. cardDeclaration's absent / missing split is keyed on CLAIM_COMMENT_MARKER, not on the branch parse; on the measured platform-objects(i18n): the zh-CN dashboard helpText leaf goes stale the moment #14586 lands — a fresh instance of the class #14931 names, with every gate green #15511 shape the thread lands on missing, whose sentence was accurate. The harm came from the anchor, and the anchor is fixed. check-half-states: a Branch line written as a markdown bullet parses to ZERO claimed branches, so the claim silently stops governing while the refusal message blames a missing claim comment #16170 closes on the anchor alone; no card for the governingClaim === null diagnosis until a live thread lands there.

Out-of-scope findings: #16217 (H22's closed-card pass truncated at the 12-page ceiling) is graded into this lane's queue in this stroke. The refsTargets docblock observation is accepted as the dev judged it — defensible as written, no card.

Landing regime: pure code ⇒ the seat flips ready and arms SQUASH auto-merge in this same act; the merge is read on the PR object, never inferred.

---

Generated by Claude Code