docs(pm-skills): the multi-label page truncates into a false-empty lane, and the arming echo is not the landing method - #18713

Fixes #18461

Clause-②: no

One file: .claude/skills/pm-dispatch/references/platform-readings.md, held at 466 / 466 (headroom 0,
widest line 120 B). No ceiling raised. Net line change 0: one row rewritten in place, one row added, one
payment taken in the same file. skip-changeset — .claude/** is shipped by no package's files[], so
nothing published moves.

Reserved rows are untouched and verified by content: :10–:12 and :431 (#18469 PR-A's pending patch),
:209 (PR #18666's one-line change, awaiting the maintainer — it rides at :208 after the payment, byte
identical), and the nine rows PR #18689 adopted an hour before this branch (:28, :85, :249, :277–:278,
:291, :347, :357, :393, :410 — none appears in any hunk of this diff).

Row ② — rewritten in place at :50

Ruling reference: the maintainer's letter A on #18421, comment 5716260764 (2026-09-17T14:41Z, 「同意」).
Its operative conclusion is the one this row carries: auto_merge.merge_method is not a reading of what will
land — the landing shape is the merge queue's to decide.

before (120 B, the falsified form):

- 挂上的 auto-merge 存的方法恒为 `merge`,不论请求了什么;REST `auto_merge.merge_method` 读回 `merge`。

after (117 B):

- auto-merge 回读 `merge_method` 不恒定:同 `SQUASH` 载荷 `merge`/`squash` 皆现,⛔ 非落地方法判据。

Why 「不恒定」 and not either constant. 「恒为 merge」 is falsified by this card's four PRs; 「恒为
squash」 is falsified by the same table's first row and by the ruling's own measurement. Same endpoint
(PUT .../pulls/{n}/ccr/auto_merge), same {"merge_method":"SQUASH"} payload:

The #18421 ruling measured the same endpoint answering merge (both arming channels, MCP and CCR), and the
dispatching seat's own arms today echoed squash on PR #18689 / #18690 / #18700 and merge on earlier ones.
Four landings, two echoes, one landing shape ⇒ the echo varies and decides nothing. A row asserting either
constant would teach a seat to investigate a read-back that is simply not a signal.

Neighbours left as they are. :49 is about the bare PATCH draft bit — unrelated. :51
(「仓库 allow_merge_commit:false 时同样读回 merge」) does not restate 「恒」: it is one conditioned
measurement, and 「同样」 now attaches to merge, which the rewritten row names as one of the two observed
values. :53–:55 already carry the landing half (the branch rule decides; three carriers agreeing proves
nothing) and are untouched — the row above them no longer contradicts them.

Row ① — added at :272

after (118 B):

- 多标签页还静默截断:`totalCount` 231 而 `returned` 30,单标签同车道 24/24 ⇒ 求交读成空车道。

Placement — a declared deviation from the dispatch. The dispatch named the :247–:248 neighbourhood
(the MCP list_issues rows). The OR half of this reading is already in the file, in the 读数陷阱 cluster at
:269–:274: :269 「labels 数组是并集」, :270 「⛔ 永不读作交集」, :272 「失效全静默」, :273
「正确读法 = 整车道单标签一次读全加本地对 labels 求交」. Writing the new row at :247 would have put a second
home for the labels-OR fact six rows away from the first — the same duplication the dispatch forbids for row ②
(one row for one fact, never two). So the row is placed where the fact it compounds already lives: directly
between the silent-failure row :272 and the prescription :273, which it is the reason for. The fact
delivered is unchanged; only the address is.

What the row adds over :269–:274. Those rows carry the OR semantics and the prescription. They do not
carry the truncation, and the two compound: the returned page is a fraction of the set with nothing in the
response saying which fraction, so a local intersection taken over that page can return zero and read as
「本车道无活」. The measurement is inside the row: labels=["domain:cli","pm:queue"] answered totalCount=231
with returned=30, while the single-label query on the same lane in the same minute answered 24 / 24 —
complete, and intersectable. That is what makes :273 load-bearing rather than a style preference.

The payment — one dedup, :202+:203 become one row

Net: 3 lines changed, 466 in, 466 out.

Gate readings

node scripts/pm/check-governed-merges.mjs --test .claude/skills/pm-dispatch/references/platform-readings.md
exits 3 = GOVERNED (.claude/** ×1) — recorded as the verdict it is. This PR stays draft; no seat flips it
ready, enqueues it or arms auto-merge. The references tier applies: the skills seat's ## Contract review
record goes on this thread, and the seat lands it.

• pnpm check:pm-skill-ratchet :: exit 0 — 「platform-readings.md is 466 lines (ceiling 466; headroom 0)」,
  「widest table row is 0 bytes (pin 0; headroom 0)」; no line over 120 B (awk 'length > 120' returns nothing).
• pnpm check:skill-frame-sync :: exit 0 — 「the one declared copy of the decision frame is internally
  coherent … 4 axes … 74 markdown files scanned for undeclared copies」.
• pnpm lint (repo-wide, eslint . --no-inline-config) :: exit 0.
• grep -naP for control characters over the file: no hits. No tag-shaped fragment in either new row.

Derived gate list, with exit codes

Derived from the worktree with node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack
(no hand-fed paths; change set 1 path vs merge base 72dd95fa5), every command run, each exit code captured by
redirect-then-$?:

node scripts/check-closing-keyword-parity.mjs                       :: exit 0
node scripts/check-closing-keyword-parity.mjs --self-test           :: exit 0
node scripts/check-comment-mask-corpus.mjs                          :: exit 0
node scripts/pm/check-governed-queue-guard.mjs --self-test          :: exit 0
node scripts/pm/check-harness-current.mjs --self-test               :: exit 0
pnpm --filter @objectstack/lint run check:doc-formula-expressions   :: exit 0
pnpm check:agent-test-spelling                                      :: exit 0
pnpm check:doc-authoring                                            :: exit 0
pnpm check:driver-memory-census                                     :: exit 0
pnpm check:nul-bytes                                                :: exit 0
pnpm check:pm-governed-merges                                       :: exit 0
pnpm check:pm-half-states                                           :: exit 0
pnpm check:pm-skill-id-lint                                         :: exit 0
pnpm check:pm-skill-ratchet                                         :: exit 0
pnpm check:refd-timer-probe                                         :: exit 0
pnpm check:required-contexts                                        :: exit 0
pnpm check:skill-frame-sync                                         :: exit 0
pnpm check:watch-hint-literal                                       :: exit 0
pnpm check:pm-settings-deny-roster                                  :: exit 0   (outside the derivation; run
                                                                                 because its roster lives under
                                                                                 .claude, which this path is in)

check:doc-formula-expressions first exited 3 — PREREQUISITE NOT MET (@objectstack/formula and
@objectstack/lint not built, 「Nothing was measured」). It was re-run to a real verdict after
pnpm exec turbo run build --filter=@objectstack/formula --filter=@objectstack/lint through
scripts/pm/os-verify-lock.sh (VERDICT command-exit 0, held 1s, waited 0s); only the exit 0 is recorded above.

Reconciliation: node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --ran ran.txt →
「18 derived famil(ies) accounted for — 18 run, 0 NOT-MEASURED (a DERIVED zero — all 18 recorded an exit code
and none of them is 3)」. Outside that union: the 53 artifact-roster families, 11 wide-population families,
14 changeset-pending families and the CI-only families (--verify-required-set, check-half-states --provenance) — CI's, not this branch's, and named here so their absence is not read as a clearance.

Out of scope, reported not written

The card's third item — scripts/pm/check-expected-skips.mjs cannot run in a seat container
(ERR_MODULE_NOT_FOUND: Cannot find package 'yaml') — is not a references line and is not written here. In this
worktree, after pnpm install, [email protected] is present; the crash is a property of the container the seat runs
it from, not of the script.

Generated by Claude Code

Generated by Claude Code