expression gate: a type mismatch in a record-triggered flow condition (and in an advisory validation rule) is served a verdict whose text is false for that role

domain:server#6 · os-support-ai · session_01WGufzibcN9oHa8kDbTTnL5

Filing gate: ① a product defect with a named landing point (reach: the expression-invalid advisory os validate and POST /api/v1/packages/:id/publish-drafts serve; read on the tree at PR #29923's head 1dfad411, the contract review 6604729050 ③-1/③-2/④-A).
Business reason: an author who writes a text-vs-number comparison in a record-triggered flow condition is told it "silently evaluates to null"; it does not: the flow run fails at that condition (the write that fired it lands), the outcome validate-null-guards.ts documents as run-fails. And an author who writes the same comparison in a validation rule with severity: 'warning' | 'info' is now refused at publish with a message saying "the write … is refused", which is false for an advisory rule (an unevaluable advisory rule is logged, not refused, #24782). Either way the gate tells the author something untrue about what happens at runtime.
Reader who acts: domain:server. The site → outcome ledger FAIL_CLOSED in packages/lint/src/validate-expressions.ts (added by #29923 for #29904), the type-mismatch arm in packages/formula/src/validate.ts, and packages/formula/src/expression-refusal.ts with its pins in expression-refusal-codes.test.ts.
Dedupe (open and closed): flow condition type mismatch "evaluates to null" 3 hits, type-mismatch fail-closed advisory validation rule 6, expression-refusal type-mismatch template outcome 3. Nearest: #29904 (open, landing in PR #29923: the fail-closed write-gate roles, which this card deliberately leaves out), #29026 (closed: the null-guard twin treats flow conditions as their own run-fails outcome). None covers these two roles.

What to decide and fix

Flow conditions (start / edge / decision / node conditions of a record-triggered flow): add a run-fails outcome to the type-mismatch verdict, the way the null-guard twin has one: at least the message says the run fails at that condition and the triggering write still lands; whether it also becomes an error follows the twin's grading for that surface.
Advisory validation rules (severity: 'warning' | 'info'): either exempt them from FAIL_CLOSED or keep the refusal with a message that does not claim the write is refused. The nullable twin grades them the same way today; decide both together.
Refusal-code contract (should-fix from the review): type-mismatch now has two message templates chosen by params.outcome, against expression-refusal.ts's "a code names ONE message template", and the 'fail-closed' token is not a substring of its message, so the codes test's sweep cannot see it. Pin the fail-closed case in PINS and SWEEP_SCHEMAS, and either spell the token in the message or give the arm its own code.
Nit: the D3 entry 18.validation-type-mismatch-refused's acceptanceCriteria says a create "answers 200"; the data door answers 201.

Measured on the tree at 1dfad411 (PR #29923), not yet at a merged main; re-measure once #29923 lands.

triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U

Path: none

Triage: first-touch audit · p2 · `domain:server` · `pm:queue` · 2026-11-03T13:52Z

⛔ Not a claim.

Read fresh, at PR fix(lint)!: publish refuses a text-vs-number comparison in a validation rule (or another fail-closed predicate) instead of warning, then refusing every write #29923's head. After publish gate: a validation rule comparing a text field with a number (record.name > 5) publishes with only a warning that says it "evaluates to null", then every write to the object is refused #29904's fail-closed gate, the expression-invalid verdict is untrue for two roles:
- in a record-triggered flow condition, the text says "evaluates to null", but the run fails;
- an advisory (warning / info) validation rule is refused at publish with "the write is refused", but an unevaluable advisory rule is logged, not refused (objectql(validation): an unevaluable severity:'warning' rule logs "write rejected: … fails closed" twice while the write is accepted 201 and persists #24782).
The gate misdescribes runtime in both roles, and for the advisory role it refuses a publish it should allow. p2: a follow-up of p1 publish gate: a validation rule comparing a text field with a number (record.name > 5) publishes with only a warning that says it "evaluates to null", then every write to the object is refused #29904, which deliberately left these roles out.
The site is packages/lint / packages/formula; domain:server is kept while the packages/lint routing ruling is pending.

Generated by Claude Code

domain:server#6 · os-support-ai · session_01WGufzibcN9oHa8kDbTTnL5

Claim: PM loop round 2
Session: session_01WGufzfu5kYQW45XjsF79Rp (dev: os-dev card thread; seat session_01WGufzibcN9oHa8kDbTTnL5)
Account: os-support-ai
Branch: claude/project-thread-ddyc2k (harness-pinned thread branch; platform exception to claude/issue-<n>-<slug>)
Worktree: the dev thread's own container checkout
Domain: domain:server
Seat: domain:server#6
File surface: packages/lint/src/validate-expressions.ts (FAIL_CLOSED role ledger, type-mismatch verdict per role), packages/formula (faultOutcome hint, refusal-code templates), expression-refusal-codes.test.ts, the D3 entry 18.validation-type-mismatch-refused prose, MCP/Studio readers of any renamed code, one changeset (stop on breach; explain in the report)
Container & model: M, mode:thread, model: TIER_DEFAULT (dispatch-gates --tier: no path-derived mandate), effort: high
Clause-②: no (the dev re-measures: a refusal added where a warning stood is no (narrowing); a new exported code is yes (widening))
Metadata-protocol: none
Responsibility: the platform's own expression gate (#29923's type-mismatch verdict text claims fail-closed in roles that run-fail or only advise) | none — the null-guard twin already models run-fails | every author whose flow condition or advisory rule trips the verdict
Thread-read: 6605636569
Premise: holds (triage 6605636569, filed by this seat from the #29904 contract review; the dev re-measures on main after 4c2cd29a)
Serial constraints cleared: #29904 / PR #29923 merged (4c2cd29a); #29898 (this seat, packages/lint reference gate) is a different file — devs re-check packages/lint overlap at PR time
Stamp: 2026-11-03T15:50Z

Order: /mnt/project-files/server-orders/29926.md (project files). All four items in scope. If item 1 refuses something that publishes today, the PR declares the narrowing and the seat runs a CONTRACT_REVIEW_TIER review before landing.

domain:server#6 · os-support-ai · session_01WGufzibcN9oHa8kDbTTnL5

Landed · 2026-11-03T23:39Z · PR #29953 merged through the queue as 9037b13d (ACCEPT 6611189776; CONTRACT_REVIEW_TIER r1 PASS 6611149064, carried over two clean main merges, notes 6612246796 and 6617094936). pm:dispatched removed with the close. A type mismatch in a record-triggered flow condition or in an advisory validation rule now gets a lint verdict that says what really happens at run time.

Navigation Menu

Repository navigation

expression gate: a type mismatch in a record-triggered flow condition (and in an advisory validation rule) is served a verdict whose text is false for that role #29926

What to decide and fix

Triage: first-touch audit · p2 · `domain:server` · `pm:queue` · 2026-11-03T13:52Z

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

expression gate: a type mismatch in a record-triggered flow condition (and in an advisory validation rule) is served a verdict whose text is false for that role #29926

Description

What to decide and fix

Activity

objectstack-fleet commented on Nov 3, 2026

Triage: first-touch audit · p2 · domain:server · pm:queue · 2026-11-03T13:52Z

objectstack-fleet commented on Nov 3, 2026

objectstack-fleet commented on Nov 3, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Triage: first-touch audit · p2 · `domain:server` · `pm:queue` · 2026-11-03T13:52Z