docs(agents): say how a return-delivered durability seam is declared

[os-try-charles]

Fixes #16233

What changed

One paragraph of AGENTS.md — the "Degradation log levels" block that used to end at lines
938-943 on 24d622b94.

The old text told the author of a durability seam to "declare how it delivers instead —
FAILURE_PROPAGATION_CALLEES (repo-wide names) or the function-scoped
FAILURE_PROPAGATION_SITES". Both maps key on the name of a callee the catch reaches
(scripts/check-durability-degradation-log-level.mjs:505, :513), so a catch that delivers
its failure by returning an outcome object has nothing to put in either one — and the only
two remaining ways to green that gate are the two the gate's own header rejects: a baseline
entry for correct code, or a bolted-on logger.error.

The rewritten paragraph splits the declaration rule by how the failure leaves the catch:

• delivered by CALLING something → name that callee, as before (both maps, unchanged);
• delivered by RETURNING an outcome object → no callee to declare; never added to
  DURABILITY_CRITICAL_CALLEES or either FAILURE_PROPAGATION_* map, never baselined, never
  given a logger.error to satisfy the checker; pinned instead by its own file's test
  asserting the returned failure outcome (the failed-receipt case is the reference shape),
  with its population read from scripts/measure-return-propagating-durability-seams.mjs.

The invariant is untouched: silent data loss must be loud. The gate is untouched — its fourth
honest limitation (scripts/check-durability-degradation-log-level.mjs:79-99) already records
this boundary and already names the census script, so no line of that script needed to move.

Evidence

Census re-measured on this branch, reproducing the ruling's numbers exactly:

node scripts/measure-return-propagating-durability-seams.mjs --sites   :: exit 0
  2620 non-test source files · 2968 try/catch · 250 guarding a declared durable call
  · 21 returning a constructed failure object on every path
  MEMBERS 12   [1] gate sees today 0 · [2] write-shaped 4 · [3] durable on the merits 8
  ADJACENT 5 · EXCLUDED 4

The instrument this paragraph governs:

pnpm check:durability-log-level   :: exit 0
  ✓ durability-degradation log levels: 36 durability-critical catch seam(s), all loud,
    rethrowing or propagating to the caller (4 propagating, declared)
  ✓ read-seam invention: 68 read seam(s), none invents an unreported answer

node scripts/pm/dispatch-gates.mjs --commands derives 14 families for this change set; all 14
were run and reconciled with --ran (14 derived, 14 run, 0 UNRUN). 13 green:

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
pnpm check:agent-test-spelling                             :: exit 0
pnpm check:docs-audit-scope                                :: 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-governed-prose                               :: exit 0
pnpm check:pm-skill-id-lint                                :: exit 0
pnpm check:refd-timer-probe                                :: exit 0
pnpm check:required-contexts                               :: exit 0
pnpm check:watch-hint-literal                              :: exit 0

✅ The line ceiling — ruled, raised, and green

The maintainer ruled the budget. His words, verbatim and untranslated:

1. docs(agents): say how a return-delivered durability seam is declared #19214 / 卡 check:durability-log-level cannot represent a seam that propagates a durability failure by RETURNING an outcome object — its propagation lists are keyed on callee names #16233 —— AGENTS.md 行预算, A 批 +10(1099→1109)

Given as a live instruction to session_019srGWGCBBCBHqcDoRZpQRh and recorded on card #16233 as
comment 5750261694; it rules the line budget that the standing letter-C ruling 5716260390
left unruled when it named the text. ⇒ scripts/pm/check-skill-line-ratchet.mjs now carries
['AGENTS.md', 1109], with the raise recorded above the entry in that file's own ruled-raise
convention — measured cost, the re-measured rewrap headroom with its widths named, the ruling
verbatim, and the standing CROSS_FILE_MOVES sentence.

pnpm check:pm-skill-ratchet   :: exit 1  →  exit 0
  ✓ check-skill-line-ratchet: AGENTS.md is 1109 lines (ceiling 1109; headroom 0).
  ✓ check-skill-line-ratchet self-test: 157 cases pass
    — including “every live ruled-raise record quotes its ruling, dates it, and names a
      positive line count”, which is the gate validating this very comment block.

⚠️ Two corrections to the numbers this body carried before, both measured rather than
inherited:

• The rewritten paragraph measures 21 / 21 / 20 lines at 86 / 88 / 90 bytes — not
  21 / 20 / 20 — and as landed it has four lines above 88 bytes (89, 90, 89, 89; max 90), so it
  sits at ≤90 B, not at 88 B. The conclusion is unchanged and if anything stronger: at 88 B a
  rewrap would cost a line.
• “0 lossless rewrap headroom at three widths” holds only for widths ≤90 B. Re-measured with
  the gate's own wrapLine, the paragraph reflows to 18 lines at 100 B and 15 at 120 B (the whole
  section 88 → 69 at 120 B). ⇒ the honest justification for the raise is not that no width
  saves lines, it is that re-wrap is not legal currency here: this file's 2026-08-17 rule
  refuses re-wrap funding, and the 120-byte cap documents itself as codifying the corpus's
  existing ≤91-byte ASCII wrap rather than mandating a reflow to it.

Clause-②: no — no published accept set, public export, enum member, error code or authorable
key moves; the surface is governance prose plus one shrink-only ratchet data value.

Landing is still the maintainer's hand

node scripts/pm/check-governed-merges.mjs --pr 19214 → exit 3, GOVERNED — a human merge is
the review record for this PR (#9495 regime), AGENTS.md ×1, landing tier H; the second path,
scripts/pm/check-skill-line-ratchet.mjs, is not on the register and adds no governed path.
⇒ This PR stays draft: ⛔ no ready flip, ⛔ no enqueue, ⛔ no auto-merge by any seat. A line
budget is a budget, not a landing permission.

Verified independently by the accepting seat on head aa2e7408d (⛔ not taken from the dev's
report): two files vs origin/main — AGENTS.md +20/−10, byte-identical to the previous round, so
the ruled paragraph was not shrunk to buy lines — and check-skill-line-ratchet.mjs +29/−1,
one data value plus the comment block. origin/main was merged in twice (32d6e4775,
aa2e7408d); the previous head f2c3ebf6d is still an ancestor, so no history was rewritten on
this branch. 35 derived gate families, 35 run, 0 NOT-MEASURED, every one exit 0;
check:ratchet-remedy-authority re-read and unmoved (this script classifies excluded).

维护者速读(草稿)

改了什么

AGENTS.md 里讲「降级日志等级」的那一段,重写了最后半段。原文要求:把失败交给调用方的
catch,必须在门禁的两张表里声明「它是怎么交付的」。但那两张表的键都是被调函数的名字,而
一个用 return { ok: false, error } 把失败交还给调用方的 catch 根本没有可填的名字。新文
字按「失败是怎么离开 catch 的」分成两种:靠调用交付的照旧声明被调者;靠返回交付的明确
写清「没有名字可声明」,不进任何名单、不进 baseline、也不准为了让门禁变绿而硬加一行
logger.error,改由该文件自己的测试钉住,并给出普查脚本作为「这类缝一共有多少」的唯一读数
来源。⭐ 不变量本身没有动:静默的数据丢失必须吵。

为什么改

这是一条写在制度里却没人能照做的规则。今天不出事,只因为门禁当前看得见的那一类里这种缝是
0 个;树上真实存在的是 12 个(普查脚本现读,与裁决数字逐字一致),每一个都已由自己文件的
测试钉住,没有一个是未受保护的降级。真正的缺陷是那句话会把下一个开发者送进死路:正确的代码会
把门禁弄红,而唯二能弄绿的办法都是门禁自己明文拒绝的。本次改的是文字,仪器一行未动。

风险与代价(含回滚)

• 代价:AGENTS.md 这份受管文件的行数上限已经顶满(1099,余量 0),这一段净增 10 行。
  折行压缩买不来这 10 行(实测:旧段在 86/88/90 三种宽度下都是 10 行),所以要么由维护者裁一个
  行数预算把上限抬到 1109,要么这段文字落不了地。开发席不自行抬上限。
• 风险:低。只改治理文本,不改任何运行时代码、不改门禁判据、不改任何名单或 baseline;
  pnpm check:durability-log-level 在本分支现读仍是绿,36 条缝的判决一条没变。
• 回滚:单文件单段落,git revert 即可,零运行时影响。

席位意见

同意落地。domain:spec 座位 5(session_019srGWGCBBCBHqcDoRZpQRh,座位贴 #19357)在维护者直派
通道下复核并验收:上面每一条读数本席都在 head aa2e7408d 上独立重取过,⛔ 未采信 dev 的终报自
述;棘轮只动了 :1466 那一个数,AGENTS.md 正文一字未改。⚠️ 本 PR 的卡归 domain:devx 席
(#6023),本席只做验收与记录,⛔ 不接管该卡、⛔ 不改它的标签或 assignee。

你要做的

1. 读上面那段新文字,确认它就是裁决要的意思(尤其「返回式交付的缝由它自己文件的测试钉住」这一
   条)。
2. 裁一个行数预算 —— 已裁:A,+10(1099→1109)。上限已按该裁定抬到 1109,
   pnpm check:pm-skill-ratchet 由 exit 1 转 exit 0,裁决原话已写进棘轮注释块。此项无需再做。
3. 人工把这个 draft PR 合并 —— 受管面,人工合并即复核记录(governance: human merge IS the review record for governed surfaces — retire the per-PR ADR approval check, add a report-only post-merge audit #9495),⛔ 任何席位都不翻
   ready、不入队、不挂 auto-merge。这是本 PR 余下的唯一一步。

_{Authored by Claude Code, session session_017ef78bLdybu3AffehKkhfk (domain:devx seat, PR #19214's own branch). Body last revised by domain:spec seat 5, session_019srGWGCBBCBHqcDoRZpQRh, under the maintainer's direct-dispatch channel — the ruling record above. ⛔ No seat flips this PR to ready.}

---

Generated by Claude Code

[os-elon-musk]

os-dev-report

{
"issue": 16233,
"status": "done",
"branch": "claude/issue-16233-agents-return-propagating-seams",
"pr": "#19214",
"session": "session_019srGWGCBBCBHqcDoRZpQRh",
"premise_still_valid": true,
"summary": "Merged origin/main into the PR branch TWICE (merge commits 32d6e47 and aa2e740; never rebased, never force-pushed, never amended, so os-try-charles's checkout fast-forwards), then made the one change the maintainer's letter-A budget authorises: CEILINGS ['AGENTS.md', 1099] becomes ['AGENTS.md', 1109] in scripts/pm/check-skill-line-ratchet.mjs, with a 29-line ruled-raise comment block above it in that file's own established convention (measured cost; my OWN re-measured lossless-rewrap headroom with the widths named; the ruling verbatim and untranslated with all three provenance parts; and the standing 'AGENTS.md is not a CROSS_FILE_MOVES destination, so no ruledRaises record applies. Landed count, headroom 0, same convention.' sentence). AGENTS.md itself was NOT touched: the ruled paragraph is byte-identical to what the branch already carried (my total diff vs origin/main is AGENTS.md +20/-10, unchanged from the previous round, plus check-skill-line-ratchet.mjs +29/-1). PR #19214 stays DRAFT: no ready flip, no enqueue, no auto-merge, no label, no assignee, no body, no title, no card comment. Two things I was told measured FALSE and are detailed in out_of_scope_findings; two items need the dispatching seat's hand because a dev may not write them (the PR body, and the claim's declared file surface).",
"tests": "ALL readings on the final head aa2e740 (git rev-parse --short HEAD), worktree /home/user/wt-16233-ratchet, repo objectstack-ai/objectstack. Every exit code captured with redirect-then-$?, never across a pipe.\n\n1. COMPLETION CRITERION, both legs, on the MERGED tree:\n BEFORE (merged tree, ratchet entry still 1099): pnpm check:pm-skill-ratchet -> EXIT=1, one failing line: 'x check-skill-line-ratchet: AGENTS.md is 1109 lines; the ratchet ceiling is 1099. ... Raising a ceiling requires a maintainer ruling quoted in the PR.'\n AFTER (entry 1109): pnpm check:pm-skill-ratchet -> EXIT=0, and the same line now reads: 'check-skill-line-ratchet: AGENTS.md is 1109 lines (ceiling 1109; headroom 0).'\n That before/after pair IS the reverse verification here; a separate ablation is not applicable (the change is one data value plus comment lines, adds no guard, and leaves no test file behind).\n\n2. AGENTS.md LINE COUNT, re-measured on the merged tree, with the arithmetic:\n origin/main (be7382d, c334ba0 and the current ada7012 all agree): 1099 lines; ceiling on main 1099.\n merged tree: 1109 lines. Branch diff vs merge base: 1 file, +20/-10 = net +10.\n 1099 + 10 (ruled) = 1109 = the measured count = the new ceiling. Headroom 0.\n main never touched AGENTS.md in either merge window, so the ruled +10-over-landed-count is exact rather than assumed: git diff be7382d..origin/main -- AGENTS.md was EMPTY, and so was git diff c334ba0..origin/main -- AGENTS.md.\n\n3. MY OWN REWRAP-HEADROOM MEASUREMENT (imported the gate's own wrapLine from scripts/pm/check-skill-line-ratchet.mjs; the paragraph is AGENTS.md:934-953, 20 physical lines, 1708 logical bytes; per-line bytes 88 88 88 86 86 84 87 89 71 88 81 90 89 89 78 88 73 75 85 86, max 90, 4 lines above 88B). Greedy rewrap of the whole paragraph as one logical line:\n @86b -> 21 | @88b -> 21 | @90b -> 20 | @91b -> 20 | @100B -> 18 | @120B -> 15\n Section context (AGENTS.md:903-990, 66 plain-prose lines): byte width min 3 / median 85 / p90 89 / max 90; whole-section greedy rewrap 88 -> 90 @88b, 88 -> 88 @90b, 88 -> 79 @100B, 88 -> 69 @120B.\n VERDICT: 0 lossless-rewrap headroom at the width this section is actually typeset at (<=90B, inside the <=91-byte ASCII-prose convention the gate header's 'Why 120 bytes' section documents) - at 90B the rewrap returns exactly the 20 lines it already occupies and at 88B it would COST a line. Lines come back ONLY by re-typesetting ASCII prose ABOVE that convention, which the 120-byte cap documents itself as NOT imposing, and re-wrap funding is refused by this file's own 2026-08-17 rule in any case.\n\n4. DERIVED-GATE RECONCILIATION on the final head: node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 35 commands (change set: 2 paths, 60 changed lines +49/-11, merge base c334ba0, three-dot). Ran all 35. Reconciled with --ran: EXIT=0, verbatim verdict '35 derived family(ies) accounted for - 35 run, 0 NOT-MEASURED (a DERIVED zero - all 35 recorded an exit code and none of them is 3).' 35 derived / 35 run / 0 NOT-MEASURED / 0 UNRUN, and EVERY ONE exit 0. The heavy member: pnpm check:pm-dispatch-gates EXIT=0, 843.7s on this box, 1883 self-test cases (it exceeds the ~10-minute foreground cap, so it ran detached and I blocked on its PID with tail --pid in the foreground; nothing was left running). I ran the whole union TWICE - once at cffced2 (35/35/0, battery 833.4s, 1866 cases) and again at aa2e740 after the second merge, because main had moved dispatch-gates.mjs by +428/-14 in between; the re-derivation produced a byte-identical 35-command list (zero delta).\n Named gates, individually: pnpm check:pm-skill-ratchet EXIT=0 - pnpm check:ratchet-remedy-authority EXIT=0 - pnpm check:nul-bytes EXIT=0 - pnpm check:pm-governed-merges EXIT=0 - pnpm check:pm-governed-prose EXIT=0 - pnpm check:pm-skill-id-lint EXIT=0 - pnpm check:cross-package-test-inputs EXIT=0.\n\n5. check:ratchet-remedy-authority, CONFIRMED not assumed: the sweep's own verdict for the edited file, read by importing sweep() and querying the key, is verdict 'excluded', 0 live offers, 0 refused offers - identical before and after my edit, and its CONTROL corpus records expect 'excluded' for pm/check-skill-line-ratchet.mjs with set-equality audited both ways, so a flip would have reddened the gate. Gate line: '262 scripts swept; 15 mark the expanding remedy MAINTAINER-ONLY, 5 turn it down outright, 242 hand out no ratchet-expanding remedy.' UNMOVED.\n\n6. The edited script's own suite (owed because the diff edits a gate script): node scripts/pm/check-skill-line-ratchet.mjs --self-test EXIT=0, '157 cases pass', including 'every live ruled-raise record quotes its ruling, dates it, and names a positive line count'. No *.test.ts anywhere in the tree names this script (git grep over test files: zero hits), so the --self-test IS its suite.\n\n7. Repo-wide lint, NOT narrowed: pnpm lint (eslint . --no-inline-config) EXIT=0 in 1m39s, zero findings, whole population - so no narrowing declaration is owed for this card.\n\n8. Governed-surface reading, confirmed: node scripts/pm/check-governed-merges.mjs --pr 19214 -> EXIT=3, 'GOVERNED - a human merge is the review record for this PR (#9495 regime)', AGENTS.md x1, landing tier H, and 'paths not on the register: scripts/pm/check-skill-line-ratchet.mjs' (my second path adds no governed path), size 60 changed lines under the 5000 human-merge threshold.\n\n9. No package is affected (both paths are repo-root: AGENTS.md and scripts/pm/), so there is NO build closure and NO affected-package test/typecheck owed; no changeset (skip-changeset, and neither path publishes). Control-character self-scan beyond the gate: grep -naP over both edited files, zero hits. The pre-push gate printed 'check:commit-card-trailers: no card relation and no model identifier in the trailer pair' on both pushes.\n\n10. WHAT MAIN BROUGHT, and conflicts: NOTHING conflicted, either merge, and no os-regen-pending marker was ever written (both merges were clean auto-commits). The first merge brought 54 commits; of the files I care about it touched ONLY scripts/pm/dispatch-gates.mjs (+75/-57) - not AGENTS.md, not the ratchet script. The second brought 4 commits including #19284 'derive a gate population through composite actions' (dispatch-gates.mjs +428/-14) - again neither of my two files. Both --ran runs printed a STALE TREE warning afterwards because origin/main is a SHARED local ref a sibling's fetch advances; current origin/main is ada7012, it touches neither of my two files and still carries AGENTS.md at 1099, so the ruled arithmetic is unaffected and I stopped chasing it rather than merging a third time.",
"mcp_calls": "0 - no MCP GitHub tool was called at all; every GitHub read was a curl GET through the REST proxy.",
"api_writes": "2, both on my own report comment: POST /repos//issues/19214/comments (comment 5750531817) and one PATCH /repos/objectstack-ai/objectstack/issues/comments/5750531817 correcting this very field, because the first send described this write as blocked (an earlier tool call in this session had been refused by the permission layer with 'External System Writes', and the POST was then allowed). ROUTING NOTE: my agent contract makes GitHub the authoritative report channel and asks for the comment on the ISSUE, but this dispatch forbids posting on card #16233 (the seat owns that thread), so it is on PR #19214's thread instead - card thread, PR body and PR title all untouched; relay it to the card if the seat wants it there. Everything else was git, not REST: 2 pushes to the existing branch, no POST /pulls, no POST labels, no PATCH of any PR body, no state change.",
"open_questions": [
{
"question": "PR #19214's body now contradicts its own diff and carries no maintainer ruling - and the ratchet's own discipline is 'RAISING one requires a maintainer ruling quoted in the raising PR's body', with completeness explicitly left to the reviewer because the gate cannot see the PR. The body still says 'the dev seat did NOT touch scripts/pm/check-skill-line-ratchet.mjs; the +10 is reported as measured and the budget is the approver's to rule', and its rewrap table is one cell wrong (see the measured-false entry). I am barred from editing the body (and re-sending it would stack a second footer), so a dev cannot fix this.",
"options": [
"A - the seat edits the body: replace the 'One gate is RED' section with the landed raise, quote the ruling verbatim and untranslated with its three provenance parts (live instruction to session_019srGWGCBBCBHqcDoRZpQRh, recorded on #16233 comment 5750261694, executing 5716260390), correct the rewrap table to 21/21/20 at 86/88/90 B, and add the Clause-2 no line the body lacks",
"B - leave the body and rely on the in-tree ruled-raise comment block plus card comment 5750261694 as the quotation of record",
"C - the seat posts a comment on the PR quoting the ruling and leaves the body untouched"
],
"recommendation": "A, because the ratchet's header puts the quotation in the PR BODY specifically and names body completeness as the reviewer's check; the maintainer reads that body as the landing record for a governed surface, and it currently tells him the opposite of what the diff does. C is an acceptable minimum if the seat wants to avoid a body rewrite; B alone leaves the human-merge record contradicting the diff."
},
{
"question": "The claim of record (#16233 comment 5746126007) declares 'File surface: AGENTS.md' only (plus, conditionally, one header sentence in check-durability-degradation-log-level.mjs) and says 'stop on breach; explain in the report'. My dispatch ordered the edit to scripts/pm/check-skill-line-ratchet.mjs, which the maintainer's ruling authorises but that claim does not list. I am explaining rather than stopping, because the file is named by the authorising ruling itself; amending the claim's declared surface is a card-thread write I am forbidden to make.",
"options": [
"A - the seat amends the claim's file surface in the same round, adding scripts/pm/check-skill-line-ratchet.mjs",
"B - treat the maintainer's budget ruling as self-amending the declared surface and record nothing"
],
"recommendation": "A - the same-round amendment of the declared file surface is exactly the obligation an out-of-surface fix carries, and it costs one comment."
}
],
"out_of_scope_findings": [
"MEASURED FALSE (dispatch text): 'the previous dev measured 0 lossless rewrap headroom at three widths' is true only for widths <=90B. Reproduced with the gate's own wrapLine: the paragraph is minimal at 90B (20 lines, headroom 0) and would COST a line at 88B (21), but at 100B it reflows to 18 and at 120B - the gate's OWN per-line budget MAX_LINE_BYTES - to 15, and the whole section 88 -> 69. So 'headroom 0' is a statement about the typeset width, not about the file; the correct justification for the raise is that re-wrap is not legal currency here (this file's 2026-08-17 rule refuses re-wrap funding, and the 120-byte cap documents itself as codifying the corpus's existing <=91-byte ASCII wrap rather than mandating a reflow to it). My comment block states it that way, with the widths named, instead of inheriting the flat claim.",
"MEASURED FALSE (PR #19214 body, minor): it reports the rewritten paragraph as '21 / 20 / 20' at 86/88/90 bytes and says it 'is landed at the 88-byte width the surrounding block uses'. I measure 21 / 21 / 20, and the paragraph as landed has 4 lines above 88 bytes (89, 90, 89, 89; max 90), so it is landed at <=90B, not 88B. The conclusion is unaffected and if anything stronger (at 88B a rewrap costs a line). Their old-paragraph reading reproduces exactly: 10 lines at 86/88/90.",
"MEASURED FALSE (dispatch hard rule 9, and the harness attribution reminder): the prescribed trailer 'Co-Authored-By: Claude Opus 5 (noreply@anthropic.com)' is mechanically REFUSED by this repo - scripts/check-commit-card-trailers.mjs coauthorModelIdentifier() flags any harness-address co-author whose display name is not exactly 'Claude', and .githooks/pre-push runs it, so that spelling would have blocked the push. AGENTS.md rules the pair model-free. I used the repo's pair ('Claude-Session: https://claude.ai/code/session_019srGWGCBBCBHqcDoRZpQRh' plus 'Co-authored-by: Claude (noreply@anthropic.com)'), and the pre-push gate confirmed it on both pushes. Flagged, not silently chosen.",
"CONFIRMED TRUE (dispatch): the ratchet entry at :1438 read ['AGENTS.md', 1099] on the PR head and the map at :1910 is the unrelated widest-table-row pin (768) - untouched. The branch was 1 file +20/-10. check:ratchet-remedy-authority classifies the edited script 'excluded' and is unmoved. check-governed-merges --pr 19214 is exit 3, 1 of 1 register path. No needs:contract-review on the PR or the card.",
"noted, not filed: the two companion sentences from 5746211232 (remedyFor() -> propagationOption still telling a FAILING author to do what the new AGENTS.md text forbids, and the ':101-105 are all open' sentence that letter C falsified) are NOT in this PR and I did not implement them. Ruling 5716260390 closed four sibling cards on this gate not_planned and ruled 'the gate gets no further investment; if a real incident ... is ever traced to one of these gaps, that incident is the card', so filing is refused by the ruling itself - which is also why the previous seat withdrew its own filing promise. Successor: NONE - no open PR or queued card touches scripts/check-durability-degradation-log-level.mjs, so nobody is currently slated to read that file; the cost stays live and unowned until an incident or a maintainer ruling names it.",
"noted, not filed: this card's own thread (5746910798) already reports the structural property that 31 of 32 ratchet-capped files stand at headroom 0, so any ruling that writes text into a governed instruction file structurally needs a second ruling for its lines. Not mine to file and already on the maintainer's desk; re-measured incidentally here only for AGENTS.md (headroom 0 before, headroom 0 after).",
"noted, not filed: PR #19214's body carries no 'Clause-2:' line at all. The live carrier is the skip-changeset LABEL, which check-changeset-no-major.mjs reads from the pull_request payload alongside the body, and its header records a measured precedent where the declaration was absent and the label carried it - so this is a body-completeness gap for the seat, not a red. Named here because a dev may not write the body."
]
}

---

Generated by Claude Code