autoreview

Run structured code review closeout on git diffs with validation.

182|8|Updated Apr 4, 2026
One-click install
npx skills add https://github.com/compozy/agh --skill autoreview
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: autoreview
Source: https://github.com/compozy/agh/tree/main/.agents/skills/autoreview
Command: npx skills add https://github.com/compozy/agh --skill autoreview

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill solves the problem of missing actionable issues at the end of a code change by running a structured, closeout-oriented review pass that validates findings against the real code and nearby context.

Core Features & Use Cases

  • Actionable closeout code review: Runs the bundled structured review helper to surface accepted findings and stop once the review completes cleanly.
  • Target-aware review modes: Reviews dirty local work, branch/PR diffs (optionally against the real PR base), or committed changes—without forcing an inappropriate local review mode after commits.
  • Guardrails and verification loop: Rejects speculative/unrealistic edge cases and broad rewrites, encourages reading the actual code path, and reruns tests plus the review if review-triggered fixes modify code.

Quick Start

Use the autoreview skill to review your current branch or PR against origin/main by running it in branch mode with the correct base.

Frequently Asked Questions about autoreview

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a closeout code review on a git diff before merging?

A closeout code review validates findings against the real code path by running a structured helper across your git diff. It targets local work, branch PRs, or committed changes to ensure release readiness and stop cleanly once validated.

Can I review uncommitted local changes separately from a branch PR diff?

Yes, target-aware review modes allow you to review dirty local work, branch/PR diffs against the actual PR base, or committed changes independently. This prevents forcing an inappropriate local review mode after you have already committed changes.

Does automated security audit review reject speculative edge cases?

Yes, the security audit review includes guardrails that reject speculative or unrealistic edge cases and broad rewrites. It enforces reading the actual code path to ensure structured validation only surfaces accepted, actionable findings.

What happens if review-triggered fixes modify code during release readiness validation?

When review-triggered fixes modify code during release readiness validation, the verification loop automatically reruns tests and the review process. This ensures the final code path remains clean and all accepted findings are resolved.

Do I need to force a specific engine model override for structured code validation?

No, structured code validation uses engine-preserving execution without forcing engine or model overrides. It requires YAML-defined skill activation to run the advisory review with a security perspective across your chosen engine.