You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Stale-premise check. Source PR #6142 (#6050) is the origin of the measurement and is recent; the reverse-verification numbers it reports (22 red in driver-sql, 20 red in driver-turso remote) are the evidence, and nothing on main has since changed the assertion style being critiqued.
Why promote rather than hold. This is not an observation about code that might drift — it is a verified test-methodology defect with a demonstrated false-green, and the demonstration is unusually clean: on driver-sql the majority of cases went red by throwing knex's bare Undefined binding(s), an Error with code and status both undefined. So a rejection test written as expect(...).toThrow()passes on the very driver the fix targets, because the unfixed driver already throws — just without an ADR-0112 envelope. A test that cannot fail on the defect it names is worse than no test: it reads as coverage. The driver-turso remote side is the control that makes the point airtight (that transport never throws, so all 20 went red for the right reason).
Landing site.skills/** ⇒ domain:devx per the SKILL domain table. The body asks for a clause in the os-dev / pm-dispatch verification discipline, which is exactly that surface. Concrete location, and an acceptance criterion statable in one sentence: rejection-class cases assert the error's codeandstatus, not toThrow() alone; add first-sentence assertions only where wording is itself contract (#5240).
Scope guard for whoever claims it — do not over-reach. The clause belongs in the skill text; it is not a licence to rewrite existing suites in the same PR. Retrofitting live rejection tests across the driver families is separate work with its own serialization constraints, and bundling it would turn a discipline change into a repo-wide diff.
Dedup (full enumeration; search API unavailable to this seat).#6142/#6050 are the origin, #1116/#4436 the historical envelope-inventory cards, #5240 the wording-consistency rule this clause defers to. None of them state the assertion-floor rule; no duplicate.
Not target:v17 — internal verification discipline, no shipped surface.
Activity
claude commentedon Aug 7, 2026
Findings triage: PROMOTE —
finding→pm:queue+domain:devx.Stale-premise check. Source PR #6142 (#6050) is the origin of the measurement and is recent; the reverse-verification numbers it reports (22 red in
driver-sql, 20 red indriver-tursoremote) are the evidence, and nothing onmainhas since changed the assertion style being critiqued.Why promote rather than hold. This is not an observation about code that might drift — it is a verified test-methodology defect with a demonstrated false-green, and the demonstration is unusually clean: on
driver-sqlthe majority of cases went red by throwing knex's bareUndefined binding(s), an Error withcodeandstatusbothundefined. So a rejection test written asexpect(...).toThrow()passes on the very driver the fix targets, because the unfixed driver already throws — just without an ADR-0112 envelope. A test that cannot fail on the defect it names is worse than no test: it reads as coverage. Thedriver-tursoremote side is the control that makes the point airtight (that transport never throws, so all 20 went red for the right reason).Landing site.
skills/**⇒domain:devxper the SKILL domain table. The body asks for a clause in the os-dev / pm-dispatch verification discipline, which is exactly that surface. Concrete location, and an acceptance criterion statable in one sentence: rejection-class cases assert the error'scodeandstatus, nottoThrow()alone; add first-sentence assertions only where wording is itself contract (#5240).Scope guard for whoever claims it — do not over-reach. The clause belongs in the skill text; it is not a licence to rewrite existing suites in the same PR. Retrofitting live rejection tests across the driver families is separate work with its own serialization constraints, and bundling it would turn a discipline change into a repo-wide diff.
Dedup (full enumeration; search API unavailable to this seat). #6142/#6050 are the origin, #1116/#4436 the historical envelope-inventory cards, #5240 the wording-consistency rule this clause defers to. None of them state the assertion-floor rule; no duplicate.
Not
target:v17— internal verification discipline, no shipped surface.本评论来自分诊座位 Routine(#5474 试点),不构成认领。
Generated by Claude Code
hotlong commentedon Aug 7, 2026
⛔ 认领(claim)
session_01BDmDsu2575gDxeMCxXhDE3(devx 车道执行座位 PM)claude/issue-6144-rejection-assert-envelope.claude/agents/os-dev.md、.claude/skills/pm-dispatch/SKILL.md、skills/objectstack-pm-dispatch/SKILL.md(按各文件既有的验证纪律章节落点;⛔ 不触碰 frame-sync 的三轴结构段)。与在飞 lint(devx): view/page 谓词(visibleWhen/visibleOn 等)的裸标识符构建期静态校验 —— #5149 裁决拆单 #6128(packages/lint)不相交;[skill] pm-dispatch: record two meta-judgments approved 2026-08-07 (silently-dropped declarations; two implementations of one operation) #6140(同文件家族)刻意排后串行。code+status(ADR-0112 信封);toThrow()单独使用不合格;措辞属契约时({ field: {} }(零个操作符的字段约束)在同仓有三个答案:driver-sql 组合子内 TRUE、顶层抛 INVALID_FILTER、formula/driver-memory FALSE #5240)加首句断言。两种失明机制(裸 Error 恒绿 / 从不抛的 transport)都要写进条款作为动机。undefined比较数一律拒收 —— 闸落在任何发射器/守卫之前,两个毛病同闸消灭 (#6050) #6142 的反向验证实测(22+20 红的两种机制)为出处,dev 落地前重读该 PR 正文核对引用。派 1 名 os-dev。
Generated by Claude Code