Repository navigation
metadata: a draft save accepts an object name without the package namespace prefix; only publish-drafts refuses it (NAMESPACE_PREFIX) #27444
Description
domain:devx#1 · sunhaolin · session_017xRBcVkGJizGH7raJHhAvL
Filing gate: ① a product defect found by a QA run (off-clause probe, RUNNER rule 8). This is the same draft-vs-publish parity class as #26707.
Reproduction
- Boot the showcase and sign in as admin.
POST /api/v1/packageswith{"manifest":{"id":"qa.scratch.r2","name":"QA Scratch R2","version":"0.0.1","type":"app"}}. The resulting namespace isr2.PUT /api/v1/meta/object/qa_formula_pkg_r2?mode=draft&package=qa.scratch.r2with a valid object body. The name has nor2_prefix.- Expected: the draft read's
_diagnosticscarries the NAMESPACE_PREFIX finding, as Studio/metadata: a draft the editors, the draft_diagnosticsand the Problems tab all call valid is refused only at publish — and one such draft rolls back the whole package release #26707 made it carry the other publish rules. - Actual: 200
state:"draft", and the diagnostics report valid. ThenPOST /api/v1/packages/qa.scratch.r2/publish-draftsanswersoutcome:"refused"with code NAMESPACE_PREFIX. The refusal rolls back the whole batch.
QA run R10 met the same thing: badprefix_thing was staged beside two valid drafts, and publish-drafts refused all three.
Mechanism
In packages/metadata-protocol's protocol.ts, the prefix check runs only when mode === 'publish' (~L20718, "a draft is judged when it is promoted"). objectNamespacePrefixViolation( is called at ~L20721, L22997 and L23766, none of which is the draft-verdict path that #26856 added (~L7181 / ~L10769).
Expected after fix
The draft verdict equals the publish verdict for the namespace-prefix rule, so Studio's Problems view names the bad object before publish.
QA-source: #27441 · api-backend.formula-stdlib-matrix · observed beside the clauses
Generated by Claude Code
Activity
objectstack-fleet commentedon Oct 29, 2026
triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U
Path: studio | none | P①
Triage: p2 ·
domain:server·area:devpath· 2026-10-29T08:09Z⛔ Not a claim.
NAMESPACE_PREFIX). This is the same draft-versus-publish parity class as Studio/metadata: a draft the editors, the draft_diagnosticsand the Problems tab all call valid is refused only at publish — and one such draft rolls back the whole package release #26707. From the 18.3.0 RC run ([sweep] v18.3.0 release test sweep #27377 §5); it is retested on the next wave.bug,pm:queue.Generated by Claude Code
objectstack-fleet commentedon Oct 29, 2026
domain:server#2 · os-sales · session_01L6bhgUCizoYWQVJWjhdR6F
Claim: PM loop round 4
Session:
session_01L6bhgUCizoYWQVJWjhdR6FAccount:
os-salesBranch:
claude/issue-27444-draft-verdict-package-prefixWorktree:
objectstack-issue-27444Domain:
domain:serverSeat:
domain:server#2File surface:
packages/metadata-protocol/src/protocol-authoring-gate.ts(draftPublishVerdict),packages/metadata-protocol/src/protocol.ts(draft-verdict host wiring only),packages/metadata-protocol/src/protocol-save-door-refusals.ts(read; reuseobjectNamespacePrefixViolation) and their tests (stop on breach; explain in the report)Container & model:
S,mode:subagent,model: default (dispatch-gates --tier: no path-derived mandate),effort: mediumClause-②: no
Metadata-protocol: none
Responsibility:
metadata-protocol draft verdict (this lane) | #26707 made the draft verdict carry the other publish rules | every Studio author whose draft object lacks the package namespace prefixThread-read: 6491459162
Premise: holds (read on main
481ade612026-10-29T21:53Z:objectNamespacePrefixViolationis called only at publish (protocol.ts:20688), promote (:23004) and the batch pre-flight (:23787);draftPublishVerdict(protocol-authoring-gate.ts:463) builds its verdict fromjudgeRuntimeAuthoringRulesalone)Serial constraints cleared: #27670 (same
protocol.ts) merged; no open PR edits these files.objectstack-fleet commentedon Oct 29, 2026
domain:server#2 · os-sales · session_01L6bhgUCizoYWQVJWjhdR6F
os-dev-report
{
"issue": 27444,
"status": "done",
"branch": "claude/issue-27444-draft-verdict-package-prefix",
"pr": "#27780",
"head": "dda53adf",
"session": "session_01L6bhgUCizoYWQVJWjhdR6F",
"premise_still_valid": true,
"summary": "Premise measured true on 481ade6: draftPublishVerdict built its verdict from judgeRuntimeAuthoringRules alone, so an unprefixed object draft read valid:true until publish-drafts refused the batch with NAMESPACE_PREFIX. draftPublishVerdict now asks the same judge, objectNamespacePrefixViolation (the host is handed the class's own method; one check, no copy), for an object draft on the draft row's package_id (the package publish-drafts promotes it under), and prepends one issue {rule: NAMESPACE_PREFIX, path: name, message: the judge's sentence} that withPublishVerdict folds into _diagnostics as valid:false. The draft save stays ungated (200 state:draft); publish refuses as before. Clause-2 no confirmed: only a read-side diagnostic is added, nothing accepted is refused.",
"tests": "Red first: new case in packages/metadata-protocol/src/protocol.draft-publish-verdict.test.ts failed on main code (AssertionError: expected true to be false on diagnostics.valid; 1 failed | 16 passed). After fix: that file plus protocol.active-save-namespace-prefix and protocol.package-publish-audit-rows 43/43 passed. Full package
pnpm --filter @objectstack/metadata-protocol teston dda53ad: Test Files 280 passed | 4 skipped (284), Tests 21804 passed | 26 skipped; VERDICT command-exit 0. Ablation via scripts/ablation-replace.mjs: call replaced with null, anchor 1->0, blob c477eabc961a->ad467a0d1c65, test red (1 failed | 16 passed), restored blob == HEAD c477eabc961a, git diff HEAD empty. The test imports ./protocol.js relatively from src, so no dist hop and no dist preflight applies.","gates": "Local, on dda53ad (after one merge of origin/main): closure build
pnpm --filter '@objectstack/metadata-protocol^...' buildexit 0;pnpm --filter @objectstack/metadata-protocol typecheckexit 0;pnpm --filter @objectstack/metadata-protocol testexit 0;eslint --no-inline-config packages/metadata-protocolexit 0. dispatch-gates --commands: 61 derived, 60 exit 0. check:lean-entry-closure exited 3 first (objectql dist absent), exit 0 after building objectql. check:dual-build-cjs-loads exit 3, NOT MEASURED, PREREQUISITE NOT MET (needs every package's dist; CI Lint & Repo Gates runs it). --ran with exit codes: 61 derived, 60 run, 1 NOT-MEASURED, 0 UNRUN. merge-tree: none of the 16 open PRs touches the changed files. PR check-runs at report time: in_progress (Type Check, Lint, Check Changeset, Governed Surface Queue Guard and others); 3 opt-in jobs skipped.","line_budget": "4 files, +92 / -2 (changeset 11, protocol-authoring-gate.ts +20/-2, protocol.ts +6, test +55)",
"files_changed": [
".changeset/27444-draft-verdict-namespace-prefix.md",
"packages/metadata-protocol/src/protocol-authoring-gate.ts",
"packages/metadata-protocol/src/protocol.ts",
"packages/metadata-protocol/src/protocol.draft-publish-verdict.test.ts"
],
"deviations": [
"protocol-save-door-refusals.ts was not edited (read only, as the claim says); the reuse goes through the existing protocol.ts wrapper.",
"check:dual-build-cjs-loads NOT MEASURED locally (repo-wide dist prerequisite); left to CI.",
"Report comment: the seat-signature line is line 1 (post-stamped.mjs refuses otherwise) and the os-dev-report marker is line 2, so the marker is not the comment's first line as os-dev.md words it."
],
"dogfood_battery": "n/a — no
!in the PR title and no (narrowing) arm","mcp_calls": "0",
"api_writes": "2 — relay pr_create (POST /repos/objectstack-ai/objectstack/dispatches, run 40970028511, opened #27780 draft, no assignees, no reviewers); this os-dev-report comment via scripts/pm/post-stamped.mjs (POST /repos//issues/27444/comments)",
"open_questions": [],
"out_of_scope_findings": [
"carrier: none · noted, not filed: the per-item publish skips sealed managed objects before the prefix judge and the draft verdict (like the batch pre-flight) does not; whether such a draft can exist was not measured. In PR Acceptance notes."
]
}
objectstack-fleet commentedon Oct 29, 2026
domain:server#2 · os-sales · session_01L6bhgUCizoYWQVJWjhdR6F
ACCEPT → PR #27780 · head
dda53adf· 2026-10-29T23:27Z · landing sent.