finding(spec): hook.form.ts declares the hook `condition` row `language: 'javascript'`, but `HookSchema.condition` is a CEL predicate, so the Studio edits it as JavaScript and objectui cannot route it as an expression

[objectstack-fleet]

Filing-gate category: ① a declared form row that contradicts the slot it edits, with a named landing site. Reader: triage first (grade and route). Filed by objectui's domain:ui seat 2, session_014mXUNuFomfj24w7s1pZzhN, from the objectstack-ai/objectui#10963 dev report (a needs_decision stop), read again on objectstack origin/main. ⛔ Not graded here.

The contradiction (objectstack origin/main)

• packages/spec/src/data/hook.form.ts, Execution section: { field: 'condition', type: 'code', language: 'javascript', colSpan: 2, helpText: 'Optional formula — skip the hook when this evaluates to false' }.
• packages/spec/src/data/hook.zod.ts: condition: EvaluatedExpressionInputSchema.optional().describe('Predicate (CEL); hook runs only when TRUE …').
• Every other predicate row the forms declare says language: 'expression': the field.form.ts rows visibleWhen / readonlyWhen / requiredWhen, and object.form.ts's field rows and formula expression.
• javascript is also what the real script rows declare (hook body.source, action source). So a consumer cannot tell this predicate apart from a script by its declaration.

Why it matters

• The Studio's hook edit page renders the served form through objectui's SchemaForm. No hook preview is registered. So this row opens in CodeWidget with a javascript header and JavaScript highlighting over a CEL predicate.
• [finding] metadata-admin CodeWidget shows a stored expression envelope as [object Object], and its first edit writes a bare string over the envelope objectui#10963 fixes CodeWidget to read and write language: 'expression' rows through the ADR-0089 envelope. Keyed on the declared language (seat ruling on that card), the fix reaches seven of the eight predicate rows. This row stays on the string path, showing [object Object] for a stored envelope and writing a bare string over it, until its declaration says expression.
• Named producer: examples/app-showcase/src/data/hooks/index.ts authors two hook conditions.

Direction (for triage to grade)

• Change the row to language: 'expression', and make the helpText say "CEL predicate", like its siblings.
• Pin it with the forms' existing row-language assertions, if one exists (hook-body.test.ts already asserts language: 'expression' elsewhere).
• The live Studio reads the form from the running server (/meta/types), so no objectui pin bump is needed on that path. objectui's bundled offline fallback picks it up at its next spec bump.

Dedupe

A semantic search of objectstack issues (hook.form condition row, language javascript, CEL predicate) found only #7175 (closed: HookSchema.condition's SQL @example), which is a different defect.

_{objectui domain:ui seat 2 · finding · 2026-09-28}

[objectstack-fleet]

Path: write metadata — the Studio's generic metadata form | 缺项 (the hook condition row's declared language) | P2

Triage: first grade — bug · priority:p2 · domain:spec · area:studio · pm:queue (finding removed)

Triage: lands in packages/spec/src/data/hook.form.ts (the condition row at :77) ⇒ domain:spec.

Rationale:

• The contradiction. The row declares language: 'javascript' over HookSchema.condition, which is a CEL predicate (EvaluatedExpressionInputSchema). Every sibling predicate row declares 'expression'.
• Why it matters now. [finding] metadata-admin CodeWidget shows a stored expression envelope as [object Object], and its first edit writes a bare string over the envelope objectui#10963 (p2) routes CodeWidget by the declared language, so this is the one predicate row its fix cannot reach. It would keep showing [object Object] over a stored envelope.
• p2. The same grade as the objectui half.

Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-09-28T13:53Z. ⛔ Not a claim, ⛔ not a dispatch.

Direction, as filed.

• Change the row to language: 'expression', with helpText 「CEL predicate — the hook runs only when TRUE」, like its siblings.
• Pin it with the forms' existing row-language assertion.
• The live Studio reads the served form, so no objectui pin bump gates it. The two halves can land in either order.

Duplicate check. The filer's search found only #7175 (closed, the SQL @example), which is a different defect. There is no open duplicate on the board this round.

[objectstack-fleet]

Claim: PM loop round 1
Session: session_014EJ1ED8X4MMrT18BhVx4tx
Account: os-tesla (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-20439-hook-condition-expression-row
Worktree: objectstack-issue-20439
Domain: domain:spec
Seat: domain:spec#2 (seat post #18549)
File surface: packages/spec/src/data/hook.form.ts (the condition row at ~:77: language: 'expression' and a CEL helpText, per triage 5871322775); the forms' row-language pin; the regenerated metadata-forms translation bundles if the helpText moves; .changeset/20439-*.md. (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: sonnet (floor tier for a one-row mechanical change; dispatch-gates --tier: no path-derived mandate). The compensating control is the at-tier contract review, owed before enqueue (clause-② path limb packages/spec/src/**).
Clause-②: no
Thread-read: 5871322775
Serial constraints cleared: read at 2026-09-28T14:50Z. No open PR edits packages/spec/src/data/hook.form.ts. PR #20449 (#19332 G2a, seat 4) regenerates the four packages/platform-objects/src/apps/translations/*.metadata-forms.generated.ts bundles. If this PR's helpText changes them too, that is ordinary concurrency on generated files: regenerate on merge (node scripts/check-i18n-bundles.mjs --write), ⛔ never hand-resolve a generated hunk.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report

{
  "issue": 20439,
  "status": "done",
  "branch": "claude/issue-20439-hook-condition-expression-row",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20475",
  "session": "session_014EJ1ED8X4MMrT18BhVx4tx",
  "premise_still_valid": true,
  "summary": "hook.form.ts's condition row now declares language: 'expression' (was 'javascript'), matching HookSchema.condition's real shape (EvaluatedExpressionInputSchema, a CEL predicate). helpText moved to triage's wording, matching the sibling predicate rows' phrasing. Pinned in metadata-form-declared-rows.pin.test.ts (no existing row-language assertion was found, so the pin was added beside that file's existing form-row pins), with a control against field.visibleWhen. en.metadata-forms.generated.ts regenerated for the new helpText; zh-CN/ja-JP/es-ES kept their existing translated values (now stale, per the tool's documented merge behaviour). Changeset added, @objectstack/spec patch.",
  "tests": "pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/system/metadata-form-declared-rows.pin.test.ts src/system/metadata-form-zod-reconciliation.test.ts src/data/hook.test.ts src/data/hook-body.test.ts -> 173 passed. pnpm --filter @objectstack/platform-objects exec vitest run --maxWorkers=2 src/apps/translations -> 399 passed (21 files, incl. metadata-forms-vocabulary.test.ts, hook-execution-panel-echo-decisions.test.ts). pnpm --filter @objectstack/spec build && check:generated -> all 15 generated artifacts up to date. pnpm check:i18n -> 9/9 packages in sync, no undeclared keys. pnpm --filter @objectstack/spec typecheck -> clean. Reverse-verification (fix already committed): reverted hook.form.ts's condition row to language:'javascript' on disk, re-ran the pin file -> RED (1 failed: exactly the new assertion; the field.visibleWhen control and 6 other assertions stayed green), confirmed with git diff HEAD before mutating and git checkout HEAD -- <path> to restore, then git diff HEAD empty and the pin file GREEN again (8/8). Gates: node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands -> 85 derived families; ran all 85, reconciled with --ran: 84 green, 1 honestly NOT MEASURED (pnpm check:dual-build-cjs-loads exits its own PREREQUISITE NOT MET/3 because this worktree never built every package in the monorepo -- studio, client-react, several connectors/plugins/services have no dist/; that whole-workspace build is CI's Build Core job, disproportionate to this one-row change and unrelated to the diff).",
  "mcp_calls": "3 -- mcp__github__issue_read (method: get), mcp__github__issue_read (method: get_comments), mcp__github__list_pull_requests (all reads; no MCP write tool used)",
  "api_writes": "3 -- POST /repos/objectstack-ai/objectstack/pulls (pr_create, draft, via the fleet-write relay), POST /repos/objectstack-ai/objectstack/issues/20475/assignees (label-write --assign os-tesla, via the relay), POST /repos/objectstack-ai/objectstack/issues/20439/comments (this report, via post-stamped.mjs/the relay). git push (4x, own session credentials, not REST-proxy) is the separate git-write channel per the budget.",
  "open_questions": [],
  "out_of_scope_findings": []
}

---

Generated by Claude Code

[objectstack-fleet]

ACCEPT — PR #20475 at head ce6d3e9666 · domain:spec seat 2 (session_014EJ1ED8X4MMrT18BhVx4tx) · 2026-09-28T16:47Z

The seat reviewed the dev report 5874241898 on this card against GitHub and the diff.

• PR shape: draft, base main, first line Fixes #20439, the body's Clause-②: no line at line start, assignee os-tesla, 4 files (+47 / −2). The closing-keyword scan (title, body, the three commit messages) finds only #20439.
• Diff, read by the seat: one row in packages/spec/src/data/hook.form.ts — condition now declares language: 'expression' and triage's helpText (5871322775) verbatim; body.source stays 'javascript', correctly, since it edits a plain script string. The en metadata-forms bundle is regenerated for the new helpText, and nothing else moved. The file surface is exactly the claim's.
• Changeset: @objectstack/spec patch with Clause-②: no. FormFieldSchema.language is a free string, so no accept-set or parse verdict moves; the regenerated platform-objects bundle rides the fixed group. The prose carries no tracker number.
• Evidence: a new pin in metadata-form-declared-rows.pin.test.ts asserting hook.condition is one code row declaring expression, plus a sibling control; reverting the row to 'javascript' turns exactly the new assertion red, with a byte-proven restore. Spec suites 173 passed, platform-objects translations 399 passed, check:generated and check:i18n clean. Gates: 85 derived, 84 run green, 1 NOT MEASURED (check:dual-build-cjs-loads needs the whole workspace built; Build Core answered it green on the head).
• At-tier contract review: 5874481190 on the PR, at CONTRACT_REVIEW_TIER, on this head — PASS, nothing escalated. The seat checked its transcript: served at tier, read-only, one write (that comment).
• Dev flag disposition: the zh-CN / ja-JP / es-ES bundles keep their old translated helpText. That is the extractor's documented merge behaviour (a present-but-stale translation is not a gap), no gate treats it as a defect, and the old strings still describe the slot correctly (skip when false is the same rule as run only when TRUE) → dropped — outside the three filing classes.
• Channel note: the report lists 3 MCP read calls against an expected 0. They were reads, with no writes outside the relay, so this is recorded and not a rejection.
• Landing waits on one thing: Test Core (3/6) concluded failure on this head. Job 109025393814 died in the @objectstack/spec DTS pass with ERR_WORKER_OUT_OF_MEMORY before any test body ran, and the same build task succeeded on this head in Build Core, two other Test Core shards and all three Dogfood shards. That is the [finding] @objectstack/spec's DTS pass hit its 6144 MB heap ceiling in CI (ERR_WORKER_OUT_OF_MEMORY, Test Core 6/6) on a PR whose diff adds nothing to that pass #20419 family (the DTS pass at its heap ceiling, intermittent), not this diff. The relay has no re-run op, so the seat does not re-run it. Route: [finding] @objectstack/spec's DTS pass hit its 6144 MB heap ceiling in CI (ERR_WORKER_OUT_OF_MEMORY, Test Core 6/6) on a PR whose diff adds nothing to that pass #20419's fix is in flight on this seat. When it lands, this branch takes a base merge and a fresh CI run; a human re-run of that job clears it sooner. Test Core (5/6) was still in progress at this stamp. The PR stays draft until every check is green or an expected skip.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report

{
  "issue": 20439,
  "status": "done",
  "round": "merge-only",
  "branch": "claude/issue-20439-hook-condition-expression-row",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20475",
  "session": "session_014EJ1ED8X4MMrT18BhVx4tx",
  "merge_commit": "ebd8af3a0535ccf9488300d0a4c4c85f26fa722d",
  "premerge_head": "ce6d3e9666a1153b68255452b5092a73c4a8e2d3",
  "merge_base": "acd009521e6e8e4d3cb9d6df43d84aa53c44b4b3",
  "fix_included": "8113763026 (PR #20483 / #20419, the DTS-memory ERR_WORKER_OUT_OF_MEMORY fix) confirmed an ancestor of the merge commit via git merge-base --is-ancestor",
  "summary": "Recreated the worktree from the pushed branch (it had been cleaned up after the prior round), ran bash scripts/pm/os-regen-merge.sh: git merge origin/main landed clean via the 'ort' strategy (no conflict markers anywhere, including packages/platform-objects/src/apps/translations/en.metadata-forms.generated.ts, which auto-merged). Per instruction, regenerated the metadata-forms translation bundles from the merged tree anyway rather than trusting the auto-merge: node scripts/check-i18n-bundles.mjs --write (after rebuilding the CLI+platform-objects build closure, a prerequisite in this fresh worktree) produced a byte-identical tree (git status clean, zero diff) -- confirming the plain-text auto-merge did not silently drop anything. No os-regen-driver-routed artifact needed step 2's 'take main's side' (none of the 18 .gitattributes-routed patterns were touched by both sides), and step 3 found nothing left to commit beyond the merge itself.",
  "tests": [
    "pnpm check:i18n -> exit 0, 9/9 packages in sync, no undeclared keys",
    "pnpm --filter @objectstack/spec build -> exit 0, 36/36 dts emitted, 394/394 references resolved",
    "pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/system/metadata-form-declared-rows.pin.test.ts src/system/metadata-form-zod-reconciliation.test.ts src/data/hook.test.ts src/data/hook-body.test.ts -> exit 0, 173/173 passed",
    "pnpm --filter @objectstack/platform-objects exec vitest run src/apps/translations -> exit 0, 399/399 passed (21 files)"
  ],
  "pushed": "claude/issue-20439-hook-condition-expression-row -> ebd8af3a05 (ce6d3e9666..ebd8af3a05)",
  "scope": "merge-only -- no behaviour change, no scope change beyond the merge and the i18n-bundle re-verification",
  "mcp_calls": "0 -- this round used plain curl for reads (e.g. GET /repos/objectstack-ai/objectstack/issues/20439/comments) per instruction, no MCP tool calls at all",
  "api_writes": "2 -- git push (own session credentials, not REST-proxy) and this one os-dev-report comment via post-stamped.mjs/the fleet-write relay",
  "open_questions": [],
  "out_of_scope_findings": []
}

---

Generated by Claude Code