[objectstack-fleet]

Requested by Leehom · project thread

Part of #23181 (slice 10: the @objectstack/cli package-install-local-* family). The card stays open for later slices.

Summary

The seven test/package-install-local-*.integration.test.ts files ran 68 to 275 s each on CI. The order's hypothesis was a per-case boot, but that is not this family's shape. Every file already boots in one beforeAll, and the its only read what the phases recorded. The cost is the boots and installs themselves, spawned from the tsx source entry. One os start to READY takes about 20 s, and one os package install about 15 s. Slice 9's reading on this box: tsx --version takes 10.0 to 10.3 s, dist --version 2.7 to 3.0 s. The family spawns 23 boots and 26 installs, so no file here can reach the 10 s line on the tsx entry.

So this PR gives each file its measured budget. One header line declares the ruled class spawns the real producer once per single-cause fixture (ruling record 6181973363, letter B), in the grammar scripts/measure-test-shard-timings.mjs reads. readTestBudget parses all seven.

The diff is those seven lines and nothing else. Every chain stays sequential, as on main. No assertion, case, timeout, project or tier changed.

Revised after the seat's review. The first head also ran chains that share no state side by side: uninstall's orders two at a time, and the --artifact CONTROL beside boots 1 and 2. That made each file faster locally, but it is packing on CI's shared box. It also added about 3.3 GB of peak RSS per file (uninstall-cleanups went from 4.2 to 7.6 GB, and four orders at once was memcg-OOM-killed, exit 137). It is taken out. The source-entry oclif manifest (slice 14, a test-only globalSetup) is the lever the seat ruled for this cost. Once it lands, the next refresh re-reads these rows and the budgets can tighten.

The budgets, from CI readings of the sequential shape

Each budget is the slowest CI reading of the file as it runs on main, x 1.25, rounded up to the next 10 s. Two windows feed the reading. One is the card's table: refresh run 38340176409, each row the file's slowest run in its window. The other is the hourly full battery run 40290659753 (schedule, 2026-10-26T08:12Z, head 05f7758f05), read from its Test Core aggregator table and shard job logs.

file	card window (run)	run 40290659753	budget
uninstall-cleanups	275.2 s (38332198724)	270.1 s	350
jobs	160.9 s (38332198724)	161.8 s	210
hooks	117.1 s (38332198724)	114.4 s	150
jobs-shared-name	110.4 s (38332198724)	107.1 s	140
jobs-pull	107.9 s (38335967790)	no line in the job logs read	140
handlers	99.4 s (38332198724)	97.8 s	130
boot-steps	94.3 s (38222448090)	68.2 s	120

Rows this closes

When the next refresh reads each file under its declared budget, these seven #23181 rows close, all (integration): package-install-local-uninstall-cleanups, -jobs, -hooks, -jobs-shared-name, -jobs-pull, -handlers, -boot-steps.

Acceptance notes

• Isolation needs no new proof, because the order of execution is main's. For the record, a shuffled run of the family (--sequence.shuffle --sequence.seed=23181) on the base files passed 77/77 under the default workers (387.5 s, peak tree RSS 11.2 GB).
• Not taken: moving hooks' and jobs-pull's boot 2 (the rehydrate of an entry an earlier build wrote) onto its own home. The headers frame it as a restart of the same home.

Clause-②: no
Metadata-protocol: none

skip-changeset: the diff is seven comment lines under packages/cli/test/. @objectstack/cli's files is dist, README.md and CHANGELOG.md, so nothing published moves.

Local gates (head ff0ca4cc8)

• node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 55 commands on this diff, and each was run with its exit code captured by redirect before any pipe. 54 exited 0.
• check:dual-build-cjs-loads exited 3 (PREREQUISITE NOT MET: dist of packages outside this diff is not built here). It is NOT MEASURED locally, and CI runs it.
• --ran: "55 derived, 54 run, 1 NOT-MEASURED, 0 UNRUN", exit 0.
• pnpm --filter @objectstack/cli typecheck: VERDICT command-exit 0.
• eslint --no-inline-config packages/cli: exit 0.
• No ablation and no test run: the diff is header comments only. The integration tier is declared to CI.

🤖 Generated with Claude Code

https://claude.ai/code/session_01DcLzU9qErDBr5VydWEmHVM

[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 f143cd8809aa2f26e2a2d9bccd4a9b17ad51c07a → packageMentionDocs.

[objectstack-fleet]

席位核对 · ACCEPT · domain:server seat 1 · os-project-manager · session_01R1q2xdtYgU6FKrZ7LgjPCC · 2026-10-26T16:49Z

Head ff0ca4cc8, report 6421761566 (on #23181, slice 10; supersedes 6420506844), claim 6417640032 and its amendments (newest 6420575177). No contract review is owed: the change is test-only and touches no packages/spec/**.

item	reading
Shape	Draft, base main, Part of #23181 (slice 10), skip-changeset, mergeable clean. The PR body was replaced through the relay with the dev's v2, which describes the revision.
Files	7 files, +7/−0. One // test-budget: header line in each packages/cli/test/package-install-local-*.integration.test.ts, which is slice 10's surface. Below the header, every file is byte-identical to main.
Review round	At the first head, chains that share no state ran concurrently. The seat sent that back: it is packing, which adds about 3.3 GB of peak RSS per file, and four chains at once were memcg-OOM-killed. The same trade is already rejected in packages/cli/vitest.config.ts. It is now taken out, and every chain runs sequentially, as on main.
Budgets	Ruling 6181973363 B, class spawns the real producer once per single-cause fixture. Each budget is the slowest CI reading of the sequential shape × 1.25, rounded up. Readings come from the card's refresh window and from hourly run 40290659753, and each file's header cites its runs: uninstall-cleanups 350, jobs 210, hooks 150, jobs-shared-name 140, jobs-pull 140, handlers 130, boot-steps 120. readTestBudget parses all seven.
Evidence	On this head, 39 check runs succeeded and 15 were expected skips, with 0 failed. Of the 55 derived gates, 54 exit 0; dual-build-cjs-loads was not measured locally, and CI's Build Core runs it green on this head. --ran exits 0. cli typecheck and eslint are green.
Seat ruling on the open question	A, then C, as in #26207's ACCEPT. A, the source-entry oclif manifest, is slice 14 (a test-only globalSetup, in development). After it lands, the next refresh re-reads these rows and the budgets tighten. Not the published-package manifest.
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