# Pr Review

> Review pull requests or local diffs for correctness, regressions, missing tests, security, data safety, API compatibility, migration safety, performance risk, accessibility issues, operational hazards, and documentation drift. Use before merge, push, release, or whenever the user asks for code review, risk review, pre-landing review, or final check.

- Skill: `ashermahonin/pr-review` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add ashermahonin/pr-review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ashermahonin/pr-review/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Web & Frontend
- Author: ashermahonin (https://skillmd.com/u/ashermahonin)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/ashermahonin/pr-review

---


# PR Review

## Purpose

Find defects and material maintenance risks before a change lands. Review behavior first; style matters only when it obscures correctness or raises change cost.

## Change context

- Identify changed files and intent of the change.
- Find affected flows, contracts, data, security boundaries, and tests.
- Read the diff with the old behavior in mind.
- Look for missing validation and docs drift.

## Change method

1. Reconstruct the intended behavior from the request, diff, tests, and nearby contracts.
2. Trace correctness, failure handling, concurrency, compatibility, migration, and rollback effects.
3. Check authorization, secrets, privacy, untrusted input, and destructive operations where relevant.
4. Use `engineering-quality` criteria for responsibilities, dependencies, side effects, errors, and tests; do not turn preferences into findings.
5. Review performance and operations when shared or hot paths changed.
6. Report findings by severity with a tight file/line reference, concrete impact, and fix direction. If none are found, state residual risk and checks not run.

## Engineering constraints

- Keep review read-only unless the user also asks for fixes.
- Verify uncertain external behavior with current documentation before raising a finding.
- Do not approve behavior you could not reconstruct from code and evidence.

## Evidence

- Prioritized findings with evidence
- Open questions and assumptions
- Missing or unrun validation
- Residual risk and merge recommendation

## Ready when

- Findings lead the response and are ordered by impact.
- Every finding describes a plausible failure, not a hypothetical preference.
- Line references are as narrow as the evidence permits.
- A clean review still reports test gaps and residual risk.

## Handoff

Hand off actionable findings to implementation or release decision with severity and evidence.

## References

- `references/review-checklist.md`: Use this checklist for reviews.