SuperAPI AI Directory

pr-review

Agents & models

View repository

Updated 10-16 00:30

About this skill

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"
)