review · diff

git:20261018.dcbc4f7 to git:20261022.7572c8a

17 added, 22 removed. Audit A to A.

What this update changes in the checks

No check changed: every fixed check gives the same answer before and after this update.

A tripped check is a label, not a verdict: read the matched text and the lines below. This week's updates that newly trip a check.

---
name: review
description: Use when asked for a read-only review of code, documentation, configuration, or workflows. Returns evidence-backed decisions without editing.
---
# Review
- Review the requested change without editing it. The goal is a trustworthy merge decision, not the largest possible backlog.
+ Review the requested change without editing it. The goal is a trustworthy merge decision, not the longest list of findings.
- ## Evidence and scope
+ ## Evidence
- - Read the full diff against the supplied base, including cross-file relationships. When given a lens, report only its primary failure class and honor its exclusions.
- - Ground findings in inspected code, repository contracts, tests, command output, or documented framework behavior.
- - Show a reachable failure scenario. Suspicious patterns or theoretical possibilities alone are not findings.
- - Judge materiality against the supplied intent and threat model. Repository-rule drift is evidence, not automatically a blocker.
- - If an exact-head gate result was supplied, do not rerun the full gate. Use focused probes only. Instrumented probes belong in a disposable worktree; never modify the shared checkout.
+ Read the full diff against the supplied base, including how the changed files depend on each other. A finding needs a reachable failure scenario, grounded in something you inspected or ran. Judge materiality against the supplied intent and threat model. A departure from a repository rule is evidence, not automatically a blocker.
- Enumerate the surfaces owned by the lens rather than sampling them. Read other files when they prove an in-lens finding, but do not duplicate another lens’s charter.
+ Probe as much as you need, but don't modify the shared checkout. Put instrumented probes in a disposable worktree. If you were given a gate result for the exact head, don't rerun the full gate, and use focused probes instead.
- ## Decisions
+ ## Lenses
- Return exactly one decision per finding; do not add a separate severity:
+ When you're given a lens, report only its failure class and respect its exclusions. Within your lens, go through every surface it owns rather than sampling. Read files outside the lens when they prove an in-lens finding.
- - **block** — demonstrated, in intent, material under the threat model, and serious enough to stop this merge.
- - **defer** — real and actionable, but outside the PR or below the merge bar.
- - **discard** — refuted, speculative, already accepted, or too low-value to schedule.
- - **ask** — the repository cannot choose between materially different durable product or architecture outcomes, and choosing wrongly would be expensive to reverse.
+ ## Decisions
- Do not ask about inferable implementation details, reversible choices, local refactors, naming, or test shape. Prefer the smallest in-scope correction and defer optional machinery.
+ Give each finding exactly one decision:
- Documentation, comments, or instructions that the diff adds or leaves stale against the documentation rule in `AGENTS.md` are `block` findings with impact `polish`. Quote the passages to cut.
+ - `block`: demonstrated, within the intent, material under the threat model, and serious enough to stop this merge.
+ - `defer`: real and actionable, but outside this pull request or below the merge bar.
+ - `discard`: refuted, speculative, already accepted, or not worth scheduling.
+ - `ask`: the repository can't choose between materially different, durable product or architecture outcomes, and choosing wrong would be expensive to reverse.
- Also record:
+ Reserve `ask` for that case, and decide inferable or reversible choices yourself. Recommend the smallest correction within scope, and defer optional machinery.
- - `impact`: `breakage`, `weakening`, or `polish`;
- - `evidenceStatus`: `reproduced`, `demonstrated`, or `unverified`.
+ Also record each finding's `impact`: `breakage`, `weakening`, or `polish`. This is the consequence if the change ships as it is, not the size of the fix.
- Use `unverified` only for an observation outside the checkout. It normally becomes `defer`, not `block`.
+ Documentation, comments, or instructions that break the documentation rule, whether the diff adds them or leaves them stale, are `block` findings with impact `polish`. Quote the passages to cut.
## Test recommendations
- Recommend at most one minimal regression test per failure class, preferably by extending an existing or table-driven test. Do not request tests for trivial copy, static literals, type-impossible states, or examples already covered by a stronger test. Use browser tests only for browser-specific behavior, and never propose a new fixture or test mechanism for one finding.
+ Recommend at most one minimal regression test per failure class, preferably by extending an existing test. Propose browser tests only for browser-specific behavior, and never a new fixture or test mechanism for a single finding.
## Output
- When given a schema, return only schema-conformant findings and coverage. Otherwise group findings by `block`, `ask`, `defer`, and durable `discard`. If none qualify, state that and summarize the inspected surfaces and focused checks.
+ Group findings by `block`, `ask`, `defer`, and the `discard` decisions worth keeping on record. If nothing qualifies, say so and summarize the surfaces you inspected and the focused checks you ran.