test(dogfood): the cold-boot catalog file takes its database roots under tmpdir() and removes them itself

[objectstack-fleet]

Fixes #22400
Clause-②: no

What this changes

One test file, packages/qa/dogfood/test/security-catalog-cold-boot-environment-holder.dogfood.test.ts, in the private @objectstack/dogfood package:

• databaseFile() takes its mkdtempSync base from join(tmpdir(), 'catalog-cold-boot-'). Before, the base was join(process.cwd(), 'catalog-cold-boot-').
• The per-file cwd sweep no longer covers these directories, so the file records every root databaseFile() creates and removes them in a new afterAll. That hook runs after afterEach has stopped the last kernel. A refused boot leaves no kernel to stop, but its directory is still removed. This follows the harness's own BootOptions.databaseFile contract: "Callers own the file's lifetime (create it under a temp dir, delete it after)".
• The header comment now describes the new placement and why the base is spelled tmpdir(). The old text ("the files live in this test file's own working directory, which the dogfood run removes at its end") is no longer true.

Why: scripts/pm/dispatch-gates.mjs's scratch-directory scan must be able to read every mkdtempSync base. process.cwd() is not a base it reads, so the site came back UNRESOLVED and the pm dispatch-gates self-test case failed on main. This uses the same remedy as PR #21935, which fixed the same mechanism on per-file-cwd.setup.ts: the base is made readable at the site. Per triage, dispatch-gates.mjs is not touched (the guard is not loosened, and process.cwd() is not taught to the scan), and per-file-cwd.setup.ts is not touched.

Evidence

pnpm check:pm-dispatch-gates (run detached, followed with tail --pid)

tree	final line
main 83e7ae93a (unfixed, in a separate worktree pinned there)	✗ dispatch-gates self-test: 1 of 2011 case(s) failed. (the failing case: ✗ no mkdtempSync site in this tree takes a base the scan cannot read — UNRESOLVED: packages/qa/dogfood/test/security-catalog-cold-boot-environment-holder.dogfood.test.ts:108 (a base this scan cannot read: process.cwd()))
this branch d8eee2191	✓ dispatch-gates self-test: 2011 cases pass.

On this branch the case itself reads ✓ no mkdtempSync site in this tree takes a base the scan cannot read (log line 1533 in both runs). The wrapper's own battery reads ✓ check:pm-dispatch-gates --self-test: the exit contract holds in all three directions. on both trees. The unfixed run ends ELIFECYCLE Command failed with exit code 1., and the fixed run prints no such line. Battery time on this shared box: 781.7s before and 774.7s after.

The same scan, read directly through exposedScratchDirs() on both trees: sites=1625 inTree=63 exposed=0, with unresolved(mkdtempSync) going from 1 to 0 and unresolved(all) from 289 to 288. Site and in-tree counts are unchanged, so the one site moved to "outside", not out of the scan.

The dogfood file itself (through scripts/pm/os-verify-lock.sh, on d8eee2191)

pnpm --filter @objectstack/dogfood exec vitest run --maxWorkers=2 test/security-catalog-cold-boot-environment-holder.dogfood.test.ts

 Test Files  1 passed (1)
      Tests  4 passed (4)
os-verify-lock: VERDICT command-exit 0

A catalog-cold-boot-* listing under tmpdir() (/tmp here), taken before and after the run, with a poller sampling /tmp every 100 ms during it:

before (2026-10-09T04:02:10Z): count=0
seen during the run:
  2026-10-09T04:02:32.669Z /tmp/catalog-cold-boot-6eM9D3
  2026-10-09T04:02:36.183Z /tmp/catalog-cold-boot-KGxFei
  2026-10-09T04:02:37.484Z /tmp/catalog-cold-boot-eGpgIr
  2026-10-09T04:02:39.290Z /tmp/catalog-cold-boot-PzYAkQ
  (distinct seen=4, one per case)
after (2026-10-09T04:02:41Z): count=0

So the four roots really are created under tmpdir() and none survives the file. Nothing else removes them: the dogfood globalSetup teardown only sweeps its own os-dogfood-run-TAG-file- prefix.

pnpm --filter @objectstack/dogfood typecheck gives VERDICT command-exit 0. The package's tsconfig.json includes test/**/*, and tsc --noEmit --listFiles names this file once.

Derived gates

node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 53 commands from this diff (1 path vs merge base 83e7ae93a). Each was run on d8eee2191 with its exit code recorded, then reconciled with --ran:

Run reconciliation — 53 derived, 53 run, 0 NOT-MEASURED, 0 UNRUN.
✓ dispatch-gates --ran: 53 derived famil(ies) accounted for — 53 run, 0 NOT-MEASURED (a DERIVED zero — all 53 recorded an exit code and none of them is 3).

On its first pass, pnpm check:dual-build-cjs-loads exited 3 (PREREQUISITE NOT MET): eight packages outside the dogfood closure had no dist/. They were built through the lock (44 of 44 turbo tasks, all cache hits), and the re-run exited 0: ✓ check:dual-build-cjs-loads — 107 published require entry point(s) across 66 package(s) load. The record above carries that re-run's code. The other 52 exited 0 on the first pass.

ESLint, narrowed to the one changed file: eslint --no-inline-config --format json gives 1 file, 0 errors, 0 warnings, and the file is not ignored. The resolved config sets neither parserOptions.project nor projectService, so linting is not type-aware and this diff cannot change the verdict on any file it does not touch. The repo-wide pnpm lint is CI's.

Acceptance notes

• No changeset: @objectstack/dogfood is private: true and the diff is one test file.
• The full dogfood suite is not run locally. CI's Dogfood Regression Gate runs it. Locally this file ran on its own, as above.
• The workflow-valued families the derivation names as NOT MEASURED (check-shard-attestation --emit …, check-test-completeness "$RUNNER_TEMP/…", check-issue-citations --census) take a value that exists only in a CI run. They are CI's.

---

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 abd254508b861da8ed1e96c2a813541b3274fb40 → packageMentionDocs.