packages/spec: an author-facing refusal message routes to objectstack publish, a command retired with the direct-to-environment path

Found by the gate built for #12016 (scripts/check-cli-command-ids.mjs) on its first live run, and left unfixed there deliberately: packages/spec is not that card's lane, and its edit would pull a second package's regeneration family into a scripts/-only PR. Filed unassigned so the owning lane can take it with the evidence attached.

The defect

packages/spec/src/api/endpoint.zod.ts:39, inside STORED_BOOKKEEPING_GUIDANCE — the message an author meets when they hand-write a stored-envelope key onto an api declaration:

'... Remove it — publication state is managed by `objectstack publish`, not authored.'

objectstack publish resolves to no command. objectstack is a real declared bin (packages/cli/package.json declares both objectstack and os), but there is no publish command id: the registered publish ids are package publish and plugin publish.

That is not a guess. #11465 measured it against the built oclif Config and #12177 records the result in packages/cli/src/commands/package/publish.ts:

the registered publish ids are package publish and plugin publish, and no rollback id exists at all

os publish was the legacy direct-to-environment command, removed with the path that wrote sys_environment_revision.

Why it is worth fixing rather than declaring

#12177 declares two deliberate unresolvable ids in place, because those sentences are about the removal — accurate history that a rewrite would falsify. This one is the opposite shape. It is present tense and prescriptive: it tells an author, at the moment their write is refused, where publication state actually comes from. An author who follows it types a command that errors as unknown. Same class as items 1 and 2 in #11465, which that card fixed.

Correct spelling: os package publish — the message's own neighbouring sentence names publishPackage as the writer, and packages/cli/src/commands/package/publish.ts is the command that runs it.

Two more sites, same word, lower severity

Both are outside the #12016 gate's population by design (it scans non-comment string literals in source), so they will not be caught mechanically and need the same pass:

Site	Shape
`packages/spec/src/cloud/package.zod.ts:237`	a doc comment — "the `objectstack publish` CLI sets it explicitly when promoting first-party or partner content"
`content/docs/api/declarative-endpoints.mdx:91`	published prose — "Publish it (`objectstack publish`), and the two URLs answer"

The objectstack validate on that same docs line is valid (validate.ts exists, objectstack is a declared bin) — only publish is stale.

Current state

The literal is carried as a baselined violation in scripts/check-cli-command-ids.mjs, printed on every green run of pnpm check:cli-command-ids and linked to this issue. The ledger self-retires: once the string is fixed the gate reds until the entry is deleted, so this cannot rot into a silent exemption.

Back-links: #12016 (the gate) · #11465 / PR #12177 (the sibling sweep inside packages/cli, which measured the id set).

Claim: PM loop round R7, wave 2
Session: session_01E5LFCYBJ3q2s6yW6oMLxwy
Branch: claude/issue-12223-publish-command-spelling
Worktree: objectstack-issue-12223
Domain: domain:spec
File surface: packages/spec/src/api/endpoint.zod.ts (:39 refusal message), packages/spec/src/cloud/package.zod.ts (:237 doc comment), content/docs/api/declarative-endpoints.mdx (re-measured at :89 on origin/main — line drifted from the card's :91; check the page's generated/hand-written banner before editing), scripts/check-cli-command-ids.mjs (delete the self-retiring baseline entry at ~:219-226 per the gate's own design), regenerated projections if the touched describes feed any, changeset (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: opus — dispatch-gates --tier at fire: no path-derived mandate; triage recorded Clause-② no (text-only, no accept/reject change); path limb (packages/spec/src/**) still routes the PR through this seat's CONTRACT_REVIEW_TIER review at ACCEPT before enqueue
Clause-②: no
Serial constraints cleared: wave-1 PRs #12417/#12408/#12418 and wave-2 sibling #12117 (src/data/type-compat.ts) all disjoint from this surface — note #12418 touches packages/spec/scripts/** while this card touches repo-root scripts/check-cli-command-ids.mjs, different trees; H17 on-hold trigger-file index: no hit.

Generated by Claude Code

Dev claim (os-dev seat, dispatched by PM loop R7 wave 2)

Session: session_01E5LFCYBJ3q2s6yW6oMLxwy
Branch: claude/issue-12223-publish-command-spelling (pushed empty at claim time; write route verified)
Worktree: ../objectstack-issue-12223 from origin/main @ 2e24115

Scope taken: the three stale objectstack publish sites (packages/spec/src/api/endpoint.zod.ts, packages/spec/src/cloud/package.zod.ts, content/docs/api/declarative-endpoints.mdx) plus deletion of the self-retiring baselined-violation row in scripts/check-cli-command-ids.mjs. Correct spelling to be re-verified against the CLI's live registered command ids on this tree before writing. The two deliberately-historical unresolvable ids declared by #12177 are out of scope and will not be rewritten.

Generated by Claude Code

os-dev-report

{
  "issue": 12223,
  "status": "done",
  "branch": "claude/issue-12223-publish-command-spelling",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/12439",
  "premise_still_valid": true,
  "summary": "All three stale `objectstack publish` sites corrected to `os package publish`, plus deletion of the self-retiring baselined-violation row in `scripts/check-cli-command-ids.mjs`. The spelling was re-verified on this tree against the CLI's LIVE command table (built oclif Config, not docs): 61 registered ids, the only two containing `publish` are `package publish` and `plugin publish`, no bare `publish` id and no `publish` topic. Text only, no accept/reject behaviour change. All three PM mechanism assumptions held: all sites present on origin/main, the mdx line is :89, and no overlap with wave-1 PR #12418 (`packages/spec/scripts/**` vs repo-root `scripts/`). `.changeset/spec-publish-command-spelling.md` is a `patch` on `@objectstack/spec`, matching what comparable message/docstring corrections in published packages did rather than `skip-changeset`.",
  "tests": "Gate union re-run on FINAL commit `89d84d5` (git rev-parse --short HEAD from that run): 39 gates, 39 pass, 0 fail; every exit code captured before any pipe (`cmd > log 2>&1; code=$?`), family re-derived from the actual diff via `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack` (5 paths, 47 families matched) rather than from the dispatch list. LOAD-BEARING, quoting the gate's own verdict lines: before = `⚠ baselined violation — packages/spec/src/api/endpoint.zod.ts: \"objectstack publish\" (#12223)` + `✓ check-cli-command-ids: 276 command-id literal(s) across 99 file(s) ... 1 baselined violation(s) listed above).`; after = `✓ check-cli-command-ids: 277 command-id literal(s) across 99 file(s) outside packages/cli all resolve to a real command path (71 ids derived; 3 declared fixture exemptions, 0 baselined violation(s) listed above).` The 276→277 delta is the corrected literal joining the RESOLVING population; its 38-case self-test passes on both runs. Tests: `pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/api/endpoint src/cloud/package` → `Test Files 3 passed (3) / Tests 74 passed (74)`; `pnpm --filter @objectstack/metadata exec vitest run --maxWorkers=2 src/stored-envelope` → `Test Files 1 passed (1) / Tests 25 passed (25)` (the pin walking STORED_ENVELOPE_KEYS asserting each key gets this prescription). `pnpm --filter @objectstack/spec run typecheck` green, and NOT subject to the test-exclusion trap — it runs `check:test-typecheck` over `tsconfig.test.json` (`check:test-typecheck: OK — @objectstack/spec's test layer compiles`). All heavy work serialized through `scripts/pm/os-verify-lock.sh` (VERDICT command-exit 0 on each: cli build 6s, spec build 147s, lint+formula build 157s, spec tests 4s, metadata test 5s, typecheck 57s). NO ablation performed — this is a text-only change with no guard to remove. LIVE COMMAND TABLE measurement (the card's binding constraint), after `pnpm --filter @objectstack/cli build`: `oclif.bin: os | declared bins: [\"objectstack\",\"os\"]` / `plugin commandIDs (oclif-derived from the built dist/commands glob): 61` / `SANITY validate present: true` / `ids containing publish: [\"package:publish\",\"plugin:publish\"]` / `bare \"publish\" id present: false` / `topics containing publish: []`. A FIRST reading was DISCARDED rather than reported: loading `config.commands` returned `registered command ids: 4` and `findCommand('package publish'): not found` — which would have falsely killed the premise. It was broken, not a measurement: unbuilt workspace deps made oclif's per-command `import()` fail and silently drop commands, with `findCommand('validate'): not found` as the tell since validate.ts plainly exists. The reported reading uses `Plugin.commandIDs` (filename-derived by oclif, no module import) and carries the validate sanity assertion inline. NO PROJECTION TO REGENERATE, verified not assumed: these strings are refusal-guidance map values, not `.describe()` values; no generated JSON contains the literal; a full `@objectstack/spec build` left `git status` showing only the 5 intended paths; `check:authorable-surface` and `check:generated` both green. Control-byte self-scan over all 5 touched files: `grep -naP '[\\x00-\\x08\\x0b\\x0c\\x0e-\\x1f\\x7f]'` → no hits, and `pnpm check:nul-bytes` green. DECLARED NARROWING (1 gate not measured): `pnpm --filter @objectstack/spec run check:skill-examples` exits 1 with its OWN prerequisite message `packages/client-react/dist holds no .d.ts declarations — the package is not built` and its explicit warning that a verdict now would be a false green — NOT MEASURED, not a finding; this diff touches no skills/ file and CI builds the closure. Two gates failed the same PREREQUISITE-NOT-MET way (`check:doc-formula-expressions`, `check:doc-security-posture`), were re-run after building @objectstack/lint and @objectstack/formula, and are GREEN inside the 39.",
  "files_changed": [
    "packages/spec/src/api/endpoint.zod.ts",
    "packages/spec/src/cloud/package.zod.ts",
    "content/docs/api/declarative-endpoints.mdx",
    "scripts/check-cli-command-ids.mjs",
    ".changeset/spec-publish-command-spelling.md"
  ],
  "deviations": [
    "packages/spec/src/cloud/package.zod.ts: wrote \"the `os package publish` command sets it\" rather than a bare token swap that would leave \"the `os package publish` CLI\" — `os` is the CLI and `package publish` is the command, so the literal swap is ungrammatical. Still text-only.",
    "content/docs/api/declarative-endpoints.mdx: page is hand-written prose (frontmatter title/description, no generated banner), so edited directly per the dispatch's conditional — no source-and-regenerate path exists. `objectstack validate` on the same line left untouched as instructed.",
    "scripts/check-cli-command-ids.mjs: deleted the row AND replaced it with a docblock paragraph recording that the list is now empty and why an entry here cannot outlive its defect. Deletion alone would have left a docblock describing a list that no longer matches it.",
    "One declared narrowing: check:skill-examples NOT MEASURED (self-declared unbuilt prerequisite, unrelated surface) — detailed in `tests`."
  ],
  "open_questions": [],
  "out_of_scope_findings": []
}

Note on the remaining objectstack publish occurrences in the tree, deliberately left: CHANGELOG.md, docs/plans/m1-project-artifact-schema.md and docs/adr/0006-project-environment-split.v4.md. All are historical records describing the design as it stood — the same shape #12177 declares deliberate elsewhere, where a rewrite would falsify history. Not defects, so no findings filed.

Generated by Claude Code

Contract review (clause-② enqueue gate: path limb packages/spec/src/** hit; declaration limb no, confirmed from the diff — same seven keys refused on the same declarations, only the sentence changes) — PASS, and ACCEPT — reviewer of record: spec seat R7, session_01E5LFCYBJ3q2s6yW6oMLxwy, at CONTRACT_REVIEW_TIER.

Verified against the PR #12439 diff, read in full (5 files, +45/−15):

All three sites corrected to os package publish; the load-bearing measurement is the LIVE oclif command table on this tree (61 ids, only package:publish / plugin:publish contain publish, no bare id, no topic) — and the dev correctly DISCARDED a broken first reading (unbuilt deps silently dropping commands, caught by the validate sanity tell) instead of letting it falsely kill the premise. That is instrument discipline worth naming.
The ledger row deletion is the gate's own self-retiring design working (before: 1 baselined violation; after: 277 literal(s) … 0 baselined violation(s), the corrected string joining the resolving population); the replacement docblock records why the list is empty rather than describing a list that no longer matches.
Grammar-correcting the package.zod.ts swap ("CLI" → "command") and editing the hand-written mdx directly both fall inside the dispatch's own conditionals; objectstack validate untouched as instructed; cli: fix two stale os … invocations, and declare the two deliberate ones in place #12177's deliberately-historical ids untouched.
Changeset patch with two named precedents (refusal-message fix; consumer-reaching docstring fix) — right call over skip-changeset.
No governed face ⇒ normal queue path.

Landing: ready-flip + enqueue when EVERY check on PR #12439 is green (CI started 02:50Z). Card stays pm:dispatched until MERGED.

Generated by Claude Code

feat(devx): gate CLI command-id string literals outside the CLI packa…

Navigation Menu

Repository navigation

packages/spec: an author-facing refusal message routes to `objectstack publish`, a command retired with the direct-to-environment path #12223

The defect

Why it is worth fixing rather than declaring

Two more sites, same word, lower severity

Current state

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

packages/spec: an author-facing refusal message routes to objectstack publish, a command retired with the direct-to-environment path #12223

Description

The defect

Why it is worth fixing rather than declaring

Two more sites, same word, lower severity

Current state

Activity

huangyiirene commented on Aug 26, 2026

huangyiirene commented on Aug 26, 2026

huangyiirene commented on Aug 26, 2026

huangyiirene commented on Aug 26, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

packages/spec: an author-facing refusal message routes to `objectstack publish`, a command retired with the direct-to-environment path #12223