| name | pr-review |
| description | Local PR review in your session — exhaustive (5 independent review lenses with quality synthesis) or prompt-shot (single-pass five-check framework). No GitHub comment posted. Use when the user wants a local code review of a branch diff, staged changes, or PR before pushing/merging.
|
PR REVIEW: Local code review against a branch diff. Findings go to stdout — no GitHub post.
Dependency note: This skill's delegate() call requires pr-review-agent to be registered
in your session — i.e., the host app has the app-actions bundle/behavior installed (not merely
the standalone skill file). Installing only this skill via npx skill add will not, by itself,
make delegation work.
Two review approaches
Exhaustive — The agent works through 5 focused review lenses sequentially (correctness ·
architecture · patterns · tests · pedantic), each with explicit focus on only that dimension.
Findings are merged, deduplicated, and prioritized by severity before the final output.
Mirrors the structure of pipelines/pr-review-exhaustive.dot without needing the GHA attractor runtime.
Prompt-shot — Single-pass five-check review (Necessity · Layer fit · Pattern · Correctness ·
Calibration). Fast. Best for quick sanity checks or small diffs.
How to start
Tell me what to review — I'll orient and delegate to pr-review-agent:
- "Exhaustive review of my branch vs main"
- "Prompt-shot review — just the staged changes"
- "Quick review of feature/auth-refactor before I push"
- "Review the diff at HEAD vs origin/main, exhaustive"
I do not attempt review logic myself. All review work runs inside a clean-room
pr-review-agent delegate with fresh context:
delegate(
agent="pr-review-agent",
instruction="<review approach (exhaustive/prompt-shot) and diff scope>",
context_depth="none"
)