code-review-risk

name	code-review-risk
description	Review changes for bugs, regressions, missing tests, safety gaps, public/private leaks, and architecture drift before stylistic comments.
keywords	["review","bugs","regressions","tests","safety","architecture"]

code-review-risk

Review for risk first.

Procedure

1. Read the diff and the relevant tests/docs for the behavior being changed.
2. Prioritize correctness, regressions, missing validation, public/private leaks, and architecture boundary violations.
3. Check whether changes conflict with Constitution, ADRs, decision map, naming, storage, surface registry, or host ownership.
4. Ground each finding in a concrete file/line and describe the user-visible or operational impact.
5. For each confirmed finding, apply the Problem-to-Guardrail loop: name the punctual problem, general class, existing rule that should have caught it, and close with guard/test añadido, ADR/regla añadida, or deuda explícita con expiry.
6. Enforce the anti-loop rule: if closure adds 2 ciclos seguidos of ADRs, ledgers, manifests, guards, or baselines sin reducir blockers reales, stop and classify the closure as blocker directo, deuda lateral, or pendiente externo; no más gobernanza para arreglar exceso de gobernanza.
7. Separate findings from open questions and low-risk polish.
8. If no issues are found, state the remaining test gaps or residual risk.

Constraints

• Do not lead with style nits when behavioral risk exists.
• Do not assume generated or staged files are safe without inspection.
• Do not recommend reverting unrelated user work.
• Do not mark a confirmed problem simply fixed when the defect class can recur.
• Do not recommend more governance after 2 ciclos seguidos sin reducir blockers reales; classify the remaining state directly.