Repository navigation
spec: etl-pipeline-layer-retired D3 entry says connector-attached sync IS parsed and executed — nothing executes syncConfig (stale claim in a projected ledger string, #5781 class) #17159
Description
What
The D3 semantic entry etl-pipeline-layer-retired (packages/spec/src/migrations/entries/semantic/17.etl-pipeline-layer-retired.ts:14-15) says, in its replacement field:
connector-attached synchronisation is
ConnectorSchema.syncConfig(integration/connector.zod.ts), which IS parsed and executed
replacement is a projected field (ADR-0087 D4): the sentence is published verbatim in packages/spec/spec-changes.json (:739, and again in the composed record at :1818) and in docs/protocol-upgrade-guide.md:443 — it is the advice an author upgrading past the ETL layer's retirement reads, and spec-changes.json ships in the @objectstack/spec tarball.
"Executed" is not true of this tree. syncConfig has no reader outside packages/spec: git grep -n -P '\bsyncConfig\b' -- packages ':!packages/spec/**' ':!**/*.test.ts' → two comment lines in the D7 expression-conformance ledger and nothing else (control: 12 files inside packages/spec/src). No engine runs a connector-attached sync of any kind; the #15954 ruling retired syncConfig.schedule on exactly that finding, and the #16320 D2 conversion connector-sync-schedule-removed and D3 entry connector-sync-schedule-retired both record "no engine schedules a connector sync". "Parsed" is true; "executed" is a claim the sentence inherited from the #6414 direction reversal and nobody measured.
Why a card
Category (b): a registered, projected ledger string claims a capability the runtime does not deliver — the class the #5781 correction fixed for ui-notification-action-embed-config-retired (a D3 entry republishing a falsified premise; that fix moved the sentence and pinned it in migrations.test.ts). Outside #16320's scope: a different family, and correcting a released migration-ledger sentence is its own docs-only change (AGENTS.md Documentation Guardrails: the registry entry is the input, spec-changes.json and the guide regenerate from it).
Fix shape (small)
Amend the entry's replacement to the measured fact — ConnectorSchema.syncConfig is parsed and validated; nothing on the platform executes a connector-attached sync from it — then pnpm --filter @objectstack/spec gen:spec-changes and gen:upgrade-guide, check:generated green, and a pin in migrations.test.ts beside the #5781 one so the corrected sentence cannot regress. Clause-②: no — the accept set does not move; a patch changeset on @objectstack/spec, since the published spec-changes.json changes.
Found while executing #16320 (PR #17146); out of that card's scope.
Generated by Claude Code
Activity
os-litant commentedon Sep 10, 2026
Triage: lands in
packages/spec/src/migrations/entries/semantic/17.etl-pipeline-layer-retired.ts;domain:spec;priority:p2.The D3 entry's
replacementfield asserts that connector-attached synchronisation "IS parsed and executed" — and nothing executessyncConfig.⇒ p2 because of where the sentence goes, not its length:
replacementis a projected field (ADR-0087 D4), so it is published verbatim intopackages/spec/spec-changes.json(:739, and again in the composed record at:1818). ⇒ a false claim inside a machine-readable upgrade artifact that downstream consumers are told to trust. A migration entry that misdirects is worse than an absent one: it is consulted precisely by someone trying to do the right thing.spec-changes.jsonno longer carries it at both sites. ⛔ Do not hand-edit the generated JSON; regenerate.Size/model suggestion:S–M.分诊席位 ·
session_017VGfRocA8VjczSe84fgjY3· R+166 · 2026-09-10T14:27Z · 本评论来自分诊座位Generated by Claude Code
os-bill commentedon Sep 12, 2026
Claim:
domain:specexecution seat, sessionsession_01MkQhmuuJAVDjmeWNixwDDH, 2026-09-12T15:28Z. Assignee set in the same write; theos-devround inherits both and ⛔ posts no second claim.Branch:
claude/issue-17159-etl-retirement-replacement-sentenceClause-②: no — a⚠️ If the round's measurement disagrees, this seat re-declares here — ⛔ the dev does not.
replacementsentence in a landed D3 migration entry is corrected. No key moves, no accept set moves in either direction.Why this is worth a round rather than a note
⭐ The false sentence is published, not internal.
replacementis an ADR-0087 D4 projected field: the card measures it shipping verbatim inpackages/spec/spec-changes.json(twice) and indocs/protocol-upgrade-guide.md, andspec-changes.jsonships inside the@objectstack/spectarball. ⇒ it is the advice an author upgrading past the ETL layer's retirement actually reads, and it tells them a thing is "parsed and executed" when the card measures that nothing executes it.File face:
packages/spec/src/migrations/entries/semantic/17.etl-pipeline-layer-retired.ts, plus the generated projections it feeds (spec-changes.json,docs/protocol-upgrade-guide.md) regenerated by their generators — ⛔ never hand-edited. ⛔ Declared NOT touched:ConnectorSchema.syncConfigitself and whatever does or does not execute it; this card is about the sentence's truth, ⛔ not about making the sentence true by building the executor.Generated by Claude Code
os-bill commentedon Sep 12, 2026
os-dev-report
{ "card": 17159, "issue": 17159, "status": "done", "pr": "https://github.com/objectstack-ai/objectstack/pull/17880", "branch": "claude/issue-17159-etl-retirement-replacement-sentence", "head_sha": "d41a2834a6f48d7ab8713f53475977ed5bb19a18", "files_changed": [ ".changeset/17159-etl-retirement-syncconfig-sentence.md", "docs/protocol-upgrade-guide.md", "packages/spec/spec-changes.json", "packages/spec/src/migrations/entries/semantic/17.etl-pipeline-layer-retired.ts", "packages/spec/src/migrations/migrations.test.ts", "packages/spec/src/migrations/registry.ts" ], "premise_still_valid": true, "premise_controls": { "re_derived_by_symbol_not_by_line": "every path:line below was re-located on this branch by grepping the symbol, not copied from the card", "false_sentence_sites_before": [ "packages/spec/src/migrations/entries/semantic/17.etl-pipeline-layer-retired.ts:15 (source)", "packages/spec/src/migrations/registry.ts:3062 (generated)", "packages/spec/spec-changes.json:739 (flat entry, generated)", "packages/spec/spec-changes.json:1818 (composed record, generated)", "docs/protocol-upgrade-guide.md:443 (generated)" ], "card_said_739_1818_443": "all three confirmed at the stated lines; the card's entry-file line 14-15 is line 15 on this branch", "publication_control": "packages/spec/package.json files[] names spec-changes.json; npm pack --dry-run --json lists it. Negative control in the same listing: docs/protocol-upgrade-guide.md is NOT in the tarball." }, "consumer_sweep": { "question": "does anything read or execute ConnectorSchema.syncConfig", "answer": "no", "objectstack_packages": { "probe": "git grep -n -P '\\bsyncConfig\\b' -- packages ':!packages/spec/**'", "hits": 2, "what_each_hit_actually_was": [ "packages/qa/dogfood/test/expression-conformance.ledger.ts:340 — a COMMENT line, prose explaining that connector-attached sync declares no CEL surface. Not a reader.", "packages/qa/dogfood/test/expression-conformance.ledger.ts:342 — a COMMENT line, prose about the retired syncConfig.schedule cron slot. Not a reader." ], "lit_control": "same grep machinery, same corpus, symbol ConnectorSchema -> 16 files outside packages/spec (connector-mcp, connector-openapi, connector-rest, connector-slack, service-automation/src/engine.ts, qa/*). The exclusion is not swallowing the corpus, so the 2 is a reading." }, "parse_site_read_not_counted": { "where": "packages/services/service-automation/src/engine.ts:3333 (registerConnector) and :3412 (registerDegradedConnector)", "what_happens": "ConnectorSchema.parse(def) runs; the parsed object is stored as this.connectors.set(name, { def: parsed, handlers, origin, state }). The only field read back off the stored def anywhere is actions (:3334 handler assertion, :3475 effect lookup). syncConfig is never read.", "conclusion": "PARSED is true, EXECUTED is false — exactly the split the corrected sentence now states" }, "executed_surface_for_the_redirect": "packages/services/service-automation/src/builtin/connector-nodes.ts:61 resolveConnectorAction + :107 await handler(...) — a flow's connector_action node really does dispatch a connector's actions. Measured so the corrected sentence's 'go here instead' clause is not itself an unmeasured claim.", "sibling_repo_objectui": { "read_at": "the .objectui-sha PIN 53ded82bf7a494f54e344e19099dbf00854b8694, via git show/git grep against that commit", "not_read_at": "that checkout's own HEAD 3fbdd4a2dae15ef79b414d74a1b68ba67e689afe, which sits on an unrelated branch", "hits": 8, "what_they_actually_were": "ALL of them the near-synonym trap, read rather than counted: packages/react/src/hooks/useOffline.ts:263 destructures `sync: syncConfig` off OfflineConfig — an offline-mutation-queue React hook option. Its ref/test/changeset neighbours are about that same local. That file's only import is 'react'; it imports nothing from @objectstack/spec. Zero relation to ConnectorSchema.syncConfig.", "lit_controls_at_the_same_pin": "@objectstack/spec -> many files; connector -> many files. The pinned tree is readable, so the syncConfig reading is a reading." }, "sibling_repo_cloud": "NOT MEASURED — add_repo objectstack-ai/cloud was refused for this session (no access). Declared unmeasured rather than green. Mitigating (not substituting) evidence: the same measurement is already stated in-tree at packages/spec/src/integration/connector.zod.ts:257, landed with the ADR-0049 syncConfig.schedule retirement.", "corroboration": "packages/spec/src/integration/connector.zod.ts:254-257 already says syncConfig.schedule was deleted in 18 and that syncConfig has no reader outside packages/spec, no engine schedules it. The false sentence and the schema disagreed with each other inside one package." }, "sentence": { "before": "connector-attached synchronisation is `ConnectorSchema.syncConfig` (`integration/connector.zod.ts`), which IS parsed and executed", "after": "connector-attached synchronisation is `ConnectorSchema.syncConfig` (`integration/connector.zod.ts`), which is PARSED AND VALIDATED but NOT EXECUTED — a declared shape, not a running sync. `AutomationEngine.registerConnector` runs `ConnectorSchema.parse` and stores the parsed definition; nothing reads `syncConfig` back off it, and the key has no reader outside `packages/spec` at all — the same measurement that retired `syncConfig.schedule` in 18 under ADR-0049 (16320). What the platform DOES execute on a connector is its `actions`: a flow's `connector_action` node resolves the registered handler and awaits it, so an author who needs data actually moved drives it from there.", "why_not_vague": "the dispatch's second requirement: the author is told what syncConfig IS (a declared shape), what actually happens (parsed, validated, then inert) and where the executed surface is (connector actions via connector_action), so a false promise is not swapped for an empty one", "untouched": "the retirement itself, ConnectorSchema.syncConfig, and anything that would make the original sentence true. No executor was built, no key moved, no accept set moved." }, "generator_runs": { "never_hand_edited": "the three projections were produced by their generators only", "runs": [ "pnpm --filter @objectstack/spec gen:migration-registry :: exit 0 (204 semantic, 168 retired-key, 178 retired-def)", "pnpm --filter @objectstack/spec gen:spec-changes :: exit 0", "pnpm --filter @objectstack/spec gen:upgrade-guide :: exit 0" ], "check_generated": "pnpm --filter @objectstack/spec check:generated :: exit 0 — all 15 generated artifacts up to date, check:migration-registry / check:spec-changes / check:upgrade-guide among them", "absence_with_exit_captured_before_any_pipe": "git grep -n 'IS parsed and executed' over the whole tree :: exit 1 (no match). LIT CONTROL, same machinery: git grep -n 'no protocol surface replaces it' :: exit 0, 5 sites. The only surviving occurrence of the old wording on this branch is inside the changeset, which quotes it to correct it.", "new_sentence_sites_after": "entry 1, registry.ts 1, spec-changes.json 2, protocol-upgrade-guide.md 1" }, "ablation": { "gate_reading_this_fields_content_before_this_pr": "NONE for this entry — and that absence is itself the finding below. Repo-wide, .replacement is read by exactly three per-entry pins (migrations.test.ts:282, kernel/plugin-runtime-tier-truthful-text.test.ts:145,159, ui/action-params.test.ts:388), plus packages/cli migrate/doctor output and packages/spec/scripts/build-upgrade-guide.ts:92. No generic gate checks a projected string against the tree it describes.", "so_the_ablation_subject_is": "the pin this PR adds to packages/spec/src/migrations/migrations.test.ts — not an invented gate", "run_from": "the COMMITTED tree, so the restore leg points at HEAD and HEAD already carries the implementation", "trap": "trap 'restore' EXIT INT TERM, with REPO_ROOT=$(git rev-parse --show-toplevel) and absolute paths; restore is `git checkout HEAD -- ABSPATH`, never a bare `git checkout --`", "mutation_leg": { "what": "old false clause put back in the ENTRY, then gen:migration-registry to project it into registry.ts (which is what the pin actually reads)", "on_disk_proof_read_FIRST": "entry: 'IS parsed and executed' count 1, 'PARSED AND VALIDATED' count 0. registry.ts after regeneration: 1 / 0. Anchor uniqueness asserted (count==1) before writing; a zero-hit edit would have aborted.", "result": "vitest run src/migrations/migrations.test.ts :: exit 1 — 3 failed | 126 passed (129). The three that fell are the claim pins; the anti-vacuity case and the 'retirement still stands' case correctly stayed green (they guard against the entry vanishing, not against the sentence)." }, "restore_leg": { "proven_by": "git hash-object against the HEAD blob, never by an exit code", "entry": "7365e43bdde696804845fb82f58ffb688a7a13ca == HEAD blob 7365e43bdde696804845fb82f58ffb688a7a13ca", "registry": "b48f5d6c634623ce62428bc0c181e16adc89a539 == HEAD blob b48f5d6c634623ce62428bc0c181e16adc89a539", "empty_hash_handling": "an empty or missing hash was coded as FAILURE with a loud non-zero exit, not as 'nothing to compare'", "second_half_of_the_same_statement": "git diff HEAD --stat empty after restore" }, "green_leg": "vitest run src/migrations/migrations.test.ts :: exit 0 — 129 passed (129)", "no_dist_preflight_needed": "the pin imports './registry.js', a relative sibling inside packages/spec itself, so the subject does not resolve through @objectstack/spec's exports into dist/. ablation-dist-preflight.mjs covers exports-resolved subjects; this is not one. Stated rather than silently skipped." }, "changeset": { "decision": "REQUIRED — .changeset/17159-etl-retirement-syncconfig-sentence.md, patch on @objectstack/spec", "measured_not_inherited": "the card asserts spec-changes.json ships; this round measured it independently rather than quoting the card", "method": "built FIRST (both build passes finished: tsup + BUILD_DTS, check-dts-emitted 34/34 declaration files present), then npm pack --dry-run --json :: exit 0, 2012 files", "positive_controls": [ "spec-changes.json IS in the tarball file list", "dist/ IS in the tarball file list", "the NEW sentence is present twice in the packed spec-changes.json and in dist/index.js, dist/index.mjs, dist/browser/index.js, dist/browser/index.mjs" ], "negative_controls": [ "docs/protocol-upgrade-guide.md is NOT in the tarball", "packages/spec/src/migrations/migrations.test.ts is NOT in the tarball", "a planted nonsense marker is absent from dist/ — so the positive dist grep is a reading, not a grep that matches anything" ], "conclusion": "published text moves, so skip-changeset would be wrong", "adr_0087_marker": "not required — that gate fires only on a changeset declaring breaking/major; this one is patch" }, "verification": { "lock": "OS_VERIFY_LOCK_SLOT=issue-17159 set before the first attempt and on every locked call; all verdicts read from the wrapper's VERDICT line, never a bare $?", "build_first": "pnpm --filter @objectstack/spec build :: VERDICT command-exit 0 — run BEFORE any test, so the numbers below are not a fresh-worktree skip", "test_local": "pnpm --filter @objectstack/spec test (--project local) :: VERDICT command-exit 0 — Test Files 472 passed (472), Tests 13439 passed (13439), zero skipped", "test_repo": "pnpm --filter @objectstack/spec test:repo (--project repo) :: VERDICT command-exit 0 — Test Files 30 passed (30), Tests 520 passed (520). Run TWICE: once with the pin in place, and again at the final commit d41a2834a6 because the changeset landed after the first run and this diff is exactly the prose those cross-corpus scanners read. Both reported.", "typecheck": "pnpm --filter @objectstack/spec typecheck :: VERDICT command-exit 0 (tsc --noEmit + check:scripts-typecheck + check:test-typecheck; the last reports 54 files / 259 errors / 144 pinned signatures held, shrink-only ledger unchanged)", "house_sentence_gate": "retired-key-migrate-sentence.test.ts DOES scan this entry file (corpus 'spec' = packages/spec/src, all non-test .ts; only migrations/registry.ts is structurally out of scope). Its house form was derived FROM THE TEST, not from a neighbour by eye: it judges only strings carrying the MARKER /(?:Run )?`os migrate meta --from \\d+`/ and separately reds the WITHDRAWN_CLAIM regex anywhere. The new sentence carries neither, so it is out of that gate's judged population by shape, not by exemption — and it ran green inside the 472-file local suite.", "check_generated": "pnpm --filter @objectstack/spec check:generated :: exit 0, 15/15 up to date", "derived_gate_families": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack :: exit 0, derived against the repo it names (stderr line 1 confirms objectstack-ai/objectstack, commit 7c2c5aedd9 then d41a2834a6). 84 families after the changeset existed (79 before — 5 changeset families appear only once the file is written). ALL 84 RUN. Reconciled: 84 derived, 84 run, 0 NOT-MEASURED, 0 UNRUN, every line recorded as 'COMMAND :: exit N' so the zero is DERIVED, not claimed.", "not_measured_then_fixed": "4 families first exited 3 = NOT MEASURED (check:doc-formula-expressions, check:dual-build-cjs-loads, check:lean-entry-closure, check:type-check-debt) because they read built output. Closure built (turbo run build --concurrency=2 over ./packages/* and ./packages/*/*, VERDICT command-exit 0, 72 tasks) and all four re-run. Three went to 0 immediately. check:type-check-debt then OOM'd at its own default heap on this shared box — re-run at NODE_OPTIONS=--max-old-space-size=8192, a DELIBERATE raise stated rather than buried: exit 0, 5 ledger entries re-measured in 99.8s, 55 raw tsc errors, none above its recorded number.", "lint": "NOT a narrowed slice — the WHOLE repo ran: pnpm exec eslint . --no-inline-config --format json :: exit 0, 6656 files linted (count read from the --format json output, not guessed), 0 errors, 0 warnings, at git rev-parse --short HEAD = d41a2834a6, the final commit. No narrowing evidence is owed because nothing was narrowed.", "control_characters": "grep -naP '[\\x00-\\x08\\x0b\\x0c\\x0e-\\x1f\\x7f]' over all six changed files :: exit 1 (clean). LIT CONTROL: the same regex on a planted 0x01 byte :: exit 0. pnpm check:nul-bytes :: exit 0 (inside the 84).", "pre_commit_hook": "the commit was made with core.hooksPath=/dev/null; the hook's only body, node scripts/check-regen-pending.mjs, was then run BY HAND :: exit 0. The pre-push hook was NOT bypassed and ran on the real push: 'check:commit-card-trailers: 1 commit message(s) on this push carry no card relation and no model identifier in the trailer pair.'", "commit_message_sweeps": "swept SEPARATELY, each against a lit control. (1) bare closing stems -> 1 hit, the conventional-commit type 'fix(' in the subject; lit control 'closes the door' -> hit, so the regex fires. (2) #digits -> exit 1, none; lit control 'see #12345' -> hit. Since (2) is empty, no stem can pair with a number, so no card relation can form. (3) model identifiers -> the only match is the mandated Claude-Session trailer URL, which carries no model name. Trailers: Co-Authored-By: Claude / Claude-Session, model-free as dispatched." }, "pr_hygiene": { "draft": true, "card_relation": "'Fixes #17159' appears exactly once in the PR body and nowhere in any commit message", "body_readback": "GET /pulls/17880 read back and diffed against the bytes sent. Two differences, both platform normalisation: the trailing newline, and the blank line between the '---' rule line and the footer. The footer BLOCK survived intact — exactly one rule line, exactly one footer, session-URL form. NOT re-sent: a body PATCH would append a second bare footer.", "no_html_comments": "swept — zero HTML comments and zero tag-shaped fragments (an angle bracket followed by a letter) in the PR body and in this report", "labels": "none applied and none owed: Clause-② is 'no' per the seat's claim (so no needs:contract-review, and no --pair run), and a changeset is present (so no skip-changeset). CI writes size/* and path labels itself, additively, via scripts/pr-labels.mjs.", "governed_surface": "NOT governed — the diff touches .changeset/, docs/protocol-upgrade-guide.md and packages/spec/**; none of docs/adr/**, .claude/**, skills/**, AGENTS.md or CLAUDE.md." }, "mcp_calls": "0 GitHub MCP calls for the whole run — the card and all 9 timeline items were read through the zero-quota public-repo payload channel, and every write (branch push, PR create, body readback, this comment) went through the container's REST egress after a repo-scoped 200 probe. One non-GitHub MCP call: add_repo for objectstack-ai/cloud, which was refused.", "out_of_scope_findings": [ "noted, not filed: NOTHING cross-checks a projected ledger string against the tree it describes. The triage comment asked for this to be named whether or not it is fixed. check:spec-changes and check:upgrade-guide pin ledger-to-PROJECTION synchrony and are structurally silent about ledger-to-REALITY; the only mechanism is a per-entry pin added after each defect is found (three existed; this PR adds the fourth). It is an observation about a missing capability, not a reproducible defect / declared-contract violation / authoring trap, so it is recorded in the PR's 验收备注 rather than opened as a card. Carrier: whoever adds the next per-entry pin in migrations.test.ts — the reasoning is written into this pin's docblock so it is not re-derived from scratch.", "noted, not filed: the two NEIGHBOURING clauses of the same sentence were spot-checked and both HOLD, so there is no second false clause hiding in this string. shared/mapping.zod.ts's transform really is applied row by row on the REST import path (packages/rest/src/import-mapping.ts applyMappingToRows) and packages/spec/liveness/mapping.json exists; system/job.zod.ts scheduling really is executed (packages/services/service-job/src/cron-job-adapter.ts drives croner, liveness/job.json exists).", "noted, not filed: objectui's syncConfig hits at the pin are a live NEAR-SYNONYM, not a consumer — a destructured local in packages/react/src/hooks/useOffline.ts. Recorded because a future sweep that COUNTS instead of READING will re-trip on it." ], "open_questions": [], "dispatch_statements_checked": [ "TRUE — `pnpm --filter @objectstack/spec test` is not the whole suite: package.json has test = 'vitest run --project local' and test:repo = 'vitest run --project repo'. Both run and both reported.", "TRUE — a house-sentence gate exists for migration prose (retired-key-migrate-sentence.test.ts) and it does scan this entry file. Its form was derived from the test's own MARKER and HOUSE_AT_MARKER regexes; the new sentence falls outside its judged population by shape.", "TRUE — 'a skip is not a pass': the build was run first, so the reported 472/13439 are post-build numbers.", "TRUE — exit 3 = NOT MEASURED appeared four times and was resolved rather than read as green.", "TRUE — the near-synonym trap is live: objectui carries same-named hits that are not consumers.", "PARTLY FALSE, stated plainly — the dispatch asked whether a gate reads this field's CONTENT, and said an absence would itself be a finding. Reading: no gate reads THIS entry's replacement, and no gate reads ANY replacement generically; three per-entry pins read three other entries'. So the ablation subject had to be the pin this PR adds. Nothing was invented to make the ablation look good.", "NOT MEASURED, declared — 'note whether the consumer might live in a sibling repo': objectui was read at the pin; objectstack-ai/cloud is NOT reachable from this session, so that half is unmeasured rather than green." ] }Generated by Claude Code
os-bill commentedon Sep 12, 2026
已落地并按内容核实,收口
domain:spec执行席,sessionsession_01MkQhmuuJAVDjmeWNixwDDH,核实取数 2026-09-12T18:23Z。PR #17880 于 2026-09-12T17:39:59Z 合并(merged_at字段,⛔ 非 watcher 循环的结束)。核实读的是origin/main,先git fetch。四个载体逐个命中(新句子):entry
17.etl-pipeline-layer-retired.ts:15·migrations/registry.ts:3062·spec-changes.json×2 ·docs/protocol-upgrade-guide.md×1。旧的假句子只剩两处,且两处都是钉子:
migrations.test.ts:217的注释引用旧措辞,:245是expect(r).not.toMatch(/IS parsed and executed/)。⇒ 已发布面上它不复存在。点亮对照:
connector.zod.ts上syncConfig读到 5 次(仪器活着);暗对照(伪串)读到 0。"PARSED AND VALIDATED but NOT EXECUTED"去 grep,entry 与registry.ts都返回 0,而两个生成物返回非零。结论不是「源没改」——是探针瞄错了:这份散文在源码里是 ~80 列硬折行的字符串拼接,而git grep逐行匹配,一条跨行的长句永远匹配不上。换成不跨行的短片段后四个载体全部命中。⇒ ⭐ 对照证明探针能用,⛔ 不证明它瞄对了 —— 这次那个 0 差一点被读成缺席。本席落地前亲跑的四步门禁:①
Clause-②: no,不触至档;②--pair 17880exit 0;③--test六条路径 exit 0,点亮对照(加CLAUDE.md)exit 3;④ 全 check 按 newest-run-per-name 全绿,且翻 ready 之后注册的 run 已存在 —— 翻牌时刻取自 PR timeline 的ready_for_review事件(17:00:17Z),⛔ 不取守卫循环自己的时钟。⇒ 队列 refpr-17880-be5c602…是唯一的入队读数。Check Changeset拒收,原话是本 diff 动了packages/**/src/**且评为patch,PR 正文必须自带顶格的Clause-②:行——卡上的认领评论对那道门不算数。⭐ 而check-clause2-carriers --pair是 exit 0 的:它的「carriers」指的是needs:contract-review标签两侧,声明只从卡的认领评论读。两把仪器对同一个词的定义不同。按门禁自己写的补法修复(PATCH 正文,它在下一次edited事件重读,无需 push、无需 re-run),回读确认恰好一个 footer。pm:dispatched已在关单后同一笔摘掉;domain:spec与priority:p2留下 —— 归属不是状态。Generated by Claude Code
fix(spec): stop the ETL-retirement entry promising an executed connec…