[objectstack-fleet]

Requested by Leehom · project thread

Part of #23181 (slice 11). The card is a series and stays open.

Clause-②: no
Metadata-protocol: none

Test-only, comment-only diff: skip-changeset (nothing in any package's files[] changes; packages/cli/test/** is not published).

Rows this PR addresses

The ten @objectstack/cli integration files of the per-case CLI spawn family (test/generate-refuses-*.test.ts and the five named test/build-*.test.ts). Reading used: the per-file list of shard-timings refresh run 40275747937 (PR #25960, the newest refresh with a per-file list; it was not merged, and PR #26089 used the same list for the spec/lint rows). The card's own table (run 38340176409) is older and lower.

file	spawns	slowest CI reading (run)	runs over 10 s / executed	local wall, 3 runs (s)	budget declared
test/build-callable-names.test.ts	9	123.82 s (40200279306)	5 / 5	84.5 / 89.7 / 86.5	160
test/generate-refuses-unimportable-alias.test.ts	7	94.13 s (40204927516)	3 / 3	65.7 / 68.4 / 64.7	120
test/build-runtime-bundle-bodyless.test.ts	6	92.69 s (40200279306)	3 / 3	57.8 / 57.8 / 57.8	120
test/build-strict-body-count.test.ts	6	82.44 s (40204927516)	3 / 3	55.1 / 58.4 / 59.0	110
test/generate-refuses-namespace-prefix.test.ts	5	69.59 s (40204927516)	3 / 3	49.0 / 50.9 / 49.5	90
test/build-view-container-name.test.ts	5	66.67 s (40204927516)	4 / 4	49.3 / 48.8 / 46.8	90
test/generate-refuses-unparseable-name.test.ts	5	66.41 s (40200279306)	5 / 5	47.9 / 48.5 / 45.5	90
test/generate-refuses-name-outside-charset.test.ts	4	54.70 s (40204927516)	4 / 4	37.9 / 36.7 / 37.5	70
test/generate-refuses-retired-generator.test.ts	4	53.28 s (40204927516)	4 / 4	36.2 / 35.2 / 37.8	70
test/build-text-face-advisory-count.test.ts	3	41.41 s (40200279306)	5 / 5	25.3 / 25.0 / 27.2	60

Budget = slowest CI reading x 1.25, rounded up to the next 10 s (the grammar readTestBudget in scripts/measure-test-shard-timings.mjs enforces; all ten lines parse through it). Pin: the next refresh lists none of these ten in its over-budget table; each appears under "Held to the budget it declares" instead, and is listed again only if a run reads it over its own budget.

Why a declared budget, not a cheaper test

The order's hypothesis was a per-case boot that one beforeAll could share. Measured, it does not hold here:

• The cost is the spawn floor, not the cases. One tsx bin/run-dev.js spawn costs 8.2 to 8.5 s locally, and os --version alone costs 8.4 to 8.8 s: the source entry has oclif import all 72 command modules through tsx on every start, because no oclif.manifest.json exists. Every file's local wall is its spawn count x 8.5 to 9.5 s (for example 7 spawns, 65 s; 3 spawns, 25 s).
• There is nothing to share. The five generate-refuses-* files already run all their spawns in one beforeAll. Every spawn in all ten files is a different argv or a different fixture: each one is a single-cause case (a refusal, its dry-run form, the neighbouring layer's verdict, or a control). Cutting one removes a case.
• So no file with even one spawn fits the 10 s line without changing the spawn entry, which the claim leaves out. These files are exactly the class ruling record 6181973363 letter B names ("spawns the real producer once per single-cause fixture"). Each now carries its one-line // test-budget: header, as eleven other cli spawn files already do.
• Concurrency inside a file was looked at and not taken. Two parallel spawns cost the same 9.3 s as one, but four take 17.5 s and seven take 28.9 s on this 4-core box. That is packing, not less work, which is the reason packages/cli/vitest.config.ts records for rejecting maxWorkers: 4.

The lever that would make these budgets fall (a decision, not taken here)

Measured on this tree, with an oclif.manifest.json written into packages/cli for the probe and removed afterwards (git status clean): tsx bin/run-dev.js generate object order_line --dry-run drops from 8.2 to 8.5 s to 2.0 to 2.3 s, with byte-identical output (1412 bytes) and the same exit code. --version drops to 0.4 s. A refusal case (generate object class) still exits 1 in 2.2 s. Applied to this family, local wall would fall from about 520 s to about 130 s. It is reported to the seat as an open decision because it changes the spawn shape these files document.

Proof

• Diff: 10 files, +10 lines, each a header comment at line 2. No case, assertion, timeout, tier or project changes.
• Cases still isolated: shuffled run at 179c8078a, pnpm exec vitest run --project integration --maxWorkers=2 --sequence.shuffle --sequence.seed=23181 + the 10 files: Test Files 10 passed (10), Tests 91 passed (91), lock VERDICT command-exit 0.
• Wall before: the table's three local runs, under os-verify-lock, --project integration --maxWorkers=2, the default reporter's per-file module line (the number CI's per-file timings read). The after wall is the same tree's behaviour (comment-only diff), so it was not re-run three times; the shuffled run above is the after reading.
• node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack at 179c8078a: 55 commands, all exit 0 (two first exited 3 on prerequisites: check-plugin-teardown-shape --self-test on a shallow clone, fixed by fetching 621a4876; check:dual-build-cjs-loads on eight unbuilt packages, fixed by building them). --ran: "55 derived famil(ies) accounted for, 55 run, 0 NOT-MEASURED".
• pnpm --filter @objectstack/cli typecheck: exit 0 (check:test-typecheck: OK).
• pnpm exec eslint --no-inline-config packages/cli: exit 0.

Acceptance notes

• Three spawns repeat across files with the same argv in an empty project: os g object class (in unimportable-alias, unparseable-name, name-outside-charset), os g object order-line (in unimportable-alias, name-outside-charset) and os g object order_line (in unparseable-name, name-outside-charset). Each file asserts a different layer's wording on its copy, so they are not removed here. Folding them is about 4 spawns (about 35 s per run) and a cross-file reshuffle of the [Decision] Which names should os generate ACCEPT? — a gate, a sanitiser, or both, and whose charset #16726, os generate has NO name validation at all — os generate object foo.bar emits const foo.bar: Data.ServiceObject and a barrel re-export, both un-parseable TypeScript #16541 and os generate view class emits a barrel line binding a reserved word — it parses, so both the charset gate and the parse check admit it, and no consumer can import it #17410 layer pins.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

• 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.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 0 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 c7aa9cde24ca7d17563caa4e01037f3ab1b9e990 → packageMentionDocs.

[objectstack-fleet]

席位核对 · ACCEPT · domain:server seat 1 · os-project-manager · session_01R1q2xdtYgU6FKrZ7LgjPCC · 2026-10-26T15:12Z

Head 179c8078ad, report 6418998910 (on #23181, slice 11), claim 6417592580 and its amendments. No contract review is owed: the diff is comment-only.

item	reading
Shape	draft, base main, Part of #23181 (slice 11), skip-changeset, mergeable clean.
Files	10 files, +10/−0: one // test-budget: header each on the generate-refuses-* and build-* spawn files in packages/cli/test/.
Finding	The per-case-boot hypothesis does not hold for this family. Every spawn is a separate single-cause case (the generate-refuses-* files already spawn inside one beforeAll), and each tsx run-dev spawn has an 8.2–8.5 s floor. So no file can get under 10 s within the claim.
Fix	This is the ruled route for such files: ruling 6181973363 letter B ("spawns the real producer once per single-cause fixture"). Each header is sized at the slowest CI reading in refresh run 40275747937 (#25960) × 1.25, rounded up (60–160 s), and names its spawns. Departure (accepted): 10 files rather than about 8, because each change is one comment line.
Evidence	36 check runs success and 16 expected skips on this head, 0 failed. Shuffled run 10/10 files, 91/91 tests; 55/55 derived gates exit 0; eslint and typecheck clean.
Carried	The open question (the 8.5 s spawn floor caused by the missing oclif.manifest.json) is ruled A by this seat and dispatched as slice 14 (amendment 6419075788): a test-only globalSetup writes the manifest per run. A later refresh re-sizes these headers. The duplicate spawns across generate-refuses-* are named in the PR's Acceptance notes and left in place.
Landing	「按照 pm-dispatch 规则复核过的都授权」 (Leehom, 2026-10-25T11:35Z): ready and auto-merge through the relay.

---

Generated by Claude Code