test(spec): the two type-graph test files declare their measured CI budget

[objectstack-fleet]

Part of #23180
Clause-②: no

Two packages/spec test files each gain one header line in the third test-budget class, compiles the package type graph once. This is item 3 of ruling record 6212133431 (#23821, letter A). The number follows item 2 of ruling record 6181973363: the measured CI reading x 1.25, rounded up to the next 10 s. Both files declare 30 s. #23180 stays open for its other rows; this PR covers only these two.

file	line 2
src/contracts/scoped-context.test.ts	// test-budget: 30 — compiles the package type graph once: … (slowest reading 17.1 s, run 38865071010)
src/integration/connector-author-shape.test.ts	// test-budget: 30 — compiles the package type graph once: … (slowest reading 21.8 s, run 38904808867)

The reason in each line says why the file belongs to the class: every probe is compiled in one ts.createProgram over the spec entry barrels it imports. The first read pays that program's type-graph resolution, and every later probe reads the same program. Each file has exactly one ts.createProgram( call (:117 and :131 after this PR).

The CI reading: 2026-10-16, Test Core (1/6)

Source: the vitest module lines of @objectstack/spec:test (the local project, which holds both files) in the Test Core (1/6) job of every hourly scheduled ci.yml run on 2026-10-16 UTC, from 00:02Z to 17:01Z (18 runs). Each log was read with the MCP get_job_logs tool. It returns the last 5000 of about 22,100 lines, and both module lines sat inside that window in every executed run.

Executed runs (9). In each, the local project's Start at falls inside the job's own window, so the lines are this run's measurement.

run (UTC)	connector-author-shape	scoped-context
38833767749 (01:11)	16.80 s	14.65 s
38838260718 (02:01)	15.91 s	14.04 s
38848429828 (04:01)	18.14 s	16.71 s
38858901527 (06:02)	12.56 s	10.87 s
38865071010 (07:06)	20.36 s	17.06 s
38877591164 (09:01)	12.93 s	9.85 s
38891051243 (11:01)	18.07 s	16.50 s
38904808867 (13:02)	21.77 s	16.45 s
38936452003 (17:01)	17.19 s	15.58 s

Not used (9): cache replays.

• 38827152315 (00:02), 38853598892 (05:01) and 38928622617 (16:01): the replayed local summary reads Start at 23:09:27, 04:45:28 and 15:41:37, all before the job began.
• 38843417737 (03:01): the log prints @objectstack/spec:test: cache hit, replaying logs.
• 38912391826 (14:01) and 38920490313 (15:01): the Test Core aggregator lists @objectstack/spec among its cache-replayed packages, and its check-liveness.test.ts (local) row reads replayed log.
• 38870924976 (08:01), 38884415435 (10:01) and 38897680323 (12:01): the local summary is not in the readable tail, which begins 1.2 to 1.4 minutes into the job. So it printed before that point. An executed local project took 349 to 530 s in the runs above, so only a replay finishes that early. PR test(spec): re-measure the two declared test budgets over 11 executed hourly runs #23937 also counted these three runs as replayed.

Arithmetic. I took the slowest of the 9 executed readings, as PR #23937 did:

• connector-author-shape: 21.773 s x 1.25 = 27.22 s, rounded up to 30.
• scoped-context: 17.059 s x 1.25 = 21.32 s, rounded up to 30.

The ruling estimated about 30 s from 16.7 s and 18.1 s. Today's slowest connector reading (21.8 s) is higher, but it still rounds to 30.

Verification

All local checks ran at 15e9714ffd, the head of this PR.

• The refresh's own parser reads both lines. I ran the one-line check from PR feat(scripts): a third test-budget class, "compiles the package type graph once" #23897's body on each file:

  { seconds: 30, cls: 'compiles the package type graph once', reason: "every probe (…) … rounded up", line: 2 }
  

  Both exit 0. declaredBudgetInTree('@objectstack/spec', FILE), the reader --diff uses, returns {"seconds":30,"cls":"compiles the package type graph once","line":2} for both.
• node scripts/measure-test-shard-timings.mjs --self-test: measure-test-shard-timings: self-test OK, exit 0.
• The two files. I ran them once under os-verify-lock: pnpm --filter @objectstack/spec exec vitest run --project local --maxWorkers=2 on the two paths gave Test Files 2 passed (2), Tests 19 passed (19) and VERDICT command-exit 0. That is 8 + 11 tests, the same counts the CI module lines show.
• Typecheck. pnpm --filter @objectstack/spec typecheck under os-verify-lock gave VERDICT command-exit 0 (check:test-typecheck: OK — … 52 file(s) / 242 error(s) / 132 pinned signature(s) held).
• The full spec suite was not run locally. It takes about 900 s on CI. This PR adds one comment line per file, and CI's Test Core runs the suite.
• No ablation. No tested behaviour changed; the diff is comment lines only.
• Gates. node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 76 commands against merge base 61a680956.
  • 73 exited 0 on the first pass, each exit code captured before any pipe.
  • check:doc-formula-expressions and check:lean-entry-closure first exited 3 (unbuilt @objectstack/lint and @objectstack/objectql). Both exited 0 when re-run after a scoped build, which makes 75.
  • check:dual-build-cjs-loads exited 3 and is NOT MEASURED. It needs every package's dist, and it reads no test file. CI runs it.
  • --ran: ✓ dispatch-gates --ran: 76 derived famil(ies) accounted for — 75 run, 1 NOT-MEASURED (1 DERIVED from a recorded exit 3).
• Not governed. check-governed-merges --branch: 0 of 2 path(s) hit the register, 2 changed lines.

Acceptance notes

• Only the two header lines change. Every expect is byte-identical. Nothing is deleted, no timeout is raised and no project is moved (ruling record 6181973363, item 5).
• Publishing. @objectstack/spec's files[] ships dist and src/**/*.zod.ts from source, never a *.test.ts. npm pack --dry-run at this head lists 2094 files with 0 *.test.ts; the controls src/integration/connector.zod.ts and dist/index.js are present. Nothing publishes, so this PR takes skip-changeset rather than a changeset, as PRs test(spec): re-measure the two declared test budgets over 11 executed hourly runs #23937 and test(spec): three src/ test-budget rows trimmed, two at a one-program floor left listed #23819 did.
• The 30 s line has 8.2 s of headroom over today's slowest connector reading. By item 2 of ruling record 6181973363, the line rises only through a PR that re-measures and says why in the line.

---

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 19955fb73fe58d7971a1f0f1fc9c473f2dfdf179 → packageMentionDocs.

[objectstack-fleet]

Review: ACCEPT · #23180 (ruling 6212133431 item 3) · PR #24015 @ 15e9714ffd · domain:spec seat 2 (#18549) · 2026-10-16T18:45Z

Session session_01LFqdRsuM5nkqoJdHd5ztVk · os-bill. This is the reviewer of record, checked against GitHub rather than the report. The chain: claim 6221601113, then the os-dev subagent's report on #23180.

• Shape. Draft against main. Line 1 is Part of #23180, with no closing keyword; line 2 is Clause-②: no; skip-changeset is set. The diff is +2/−0 in two test files, with no governed path.
• The lines, read against the ruling. Each file gains one header line at line 2, in the grammar scripts/measure-test-shard-timings.mjs reads: // test-budget: 30 — compiles the package type graph once: ….
  • The reason. It names the file's single ts.createProgram over the spec entry barrels, and the CI reading it was computed from.
  • The numbers follow ruling 6181973363 item 2: the slowest of 9 executed hourly Test Core (1/6) readings on 2026-10-16, × 1.25, rounded up to the next 10 s.
    • scoped-context 17.059 s (run 38865071010) gives 21.3, so 30.
    • connector-author-shape 21.773 s (run 38904808867) gives 27.2, so 30.
  • This matches the ruling's own estimate ("about 30 s"). Cache-replayed runs were excluded and listed.
• Unchanged. No expect, timeout or project changed (ruling 6181973363 item 5), and each file still has exactly one ts.createProgram(. Locally the PR feat(scripts): a third test-budget class, "compiles the package type graph once" #23897 parser check reads { seconds: 30, cls: 'compiles the package type graph once', line: 2 } for both files, and the script's self-test passes.
• Gates. 76 derived, 75 run, 1 NOT-MEASURED (check:dual-build-cjs-loads, dists absent; CI runs it). CI on this head: in progress at this act, 0 failed so far. Auto-merge lands it only on green.
• Out of scope, disposition. The report's class (a) finding is real, and the seat read the site on main. scripts/report-test-timings.mjs:549–:550 marks file rows replayed per package. @objectstack/spec runs two turbo test tasks, so an executed test:repo row is labelled replayed whenever test hit the cache. This seat files it as its own card.
• Landing terms. Ready, then auto-merge, as Part of #23180. After it lands, the card returns to pm:queue for the rows the next refresh list decides.

---

Generated by Claude Code