| name | automated-pr-review |
| description | Dispatch a multi-pass, high-signal pull request review and deliver a consolidated overview plus actionable inline findings. |
Purpose
Provide a consistent, thorough PR review workflow (optimized for depth, not speed) that scales its effort to the size and complexity of the change.
When to use
- You want deeper review coverage than quick scans.
- You want repeatable review output on every PR.
- You want a single overview comment plus inline notes, while keeping final approval to a human.
Instructions
- Read the PR context
- Identify the change intent (what problem it solves).
- Identify the risk surface: auth, data integrity, migrations, concurrency, security boundaries, config changes.
- Scale review depth to change size
- Small changes: prioritize correctness, edge cases, and obvious regressions.
- Large/complex changes: do a deeper pass across architecture, invariants, and adjacent modules.
- Review for high-signal issues
- Correctness and logic bugs
- Security and permissions
- Reliability (timeouts, retries, idempotency)
- Performance pitfalls
- Observability (logs/metrics/tracing)
- Tests and coverage gaps
- Developer experience (clarity, maintainability)
- Write results in two layers
- Overview comment: a concise summary of main findings and risk areas.
- Inline findings: specific, actionable notes tied to exact locations.
- Do not approve the PR
- Provide guidance; a human still decides whether to merge.
Output format
Overview
- Intent:
- Top risks:
- Findings (highest impact first):
- Suggested follow-ups (tests, docs, rollout):
Inline findings
For each finding:
- Location:
- Issue:
- Why it matters:
- Suggested fix:
Examples
See examples/review-comment-template.md for a ready-to-copy PR review comment template.
Source
