skills.rest
jayjnu
jayjnuCOMMUNITY
Explore all skills by this author →
GitHub

auto-review

Orchestrates parallel subagent code reviews and single-writer fixes for changed Git worktrees.

Updated May 27, 2026
Search Tags:#git#code review#orchestration#auto-fix#subagent#pi extension#parallel reviewers
One-click install
npx skills add https://github.com/jayjnu/pi-auto-review --skill auto-review-jayjnu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: auto-review
Source: https://github.com/jayjnu/pi-auto-review/tree/main/skills/auto-review
Command: npx skills add https://github.com/jayjnu/pi-auto-review --skill auto-review-jayjnu

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? After an AI coding agent finishes a turn that changes files, there is no built-in quality gate to catch regressions, missing tests, or maintainability issues before the work is accepted. This Skill automates a structured review-and-fix loop so every meaningful code change gets inspected without manual prompting. ## Core Features & Use Cases - Parallel Reviewer Fanout: Dispatches baseline reviewers (correctness, tests/validation, maintainability) plus configured reviewer profiles and skills as isolated read-only subagent tasks, then synthesizes deduplicated Critical/Warning/Suggestion findings. - Single-Writer Auto-Fix: When Critical or Warning findings are accepted, exactly one fixer subagent applies the fixes, keeping the main session as orchestrator and avoiding conflicting parallel writes. - Smart Skip Gates: Skips review for clean worktrees, lockfile/generated-only changes, identical already-reviewed diffs, and committed bookkeeping follow-ups, using scoped read-only git commands. - Layered Configuration: Merges defaults, global config, and project config via a helper script, with natural-language mapping to /auto-review commands for changing settings. - Use Case: After an agent refactors a module and commits, the skill detects the committed range, fans out reviewers scoped to the worktree, synthesizes findings, and dispatches one fixer to address accepted Critical issues. ## Quick Start Ask the agent to run the auto-review workflow on the current worktree changes, for example by saying "review the changes from the last turn and fix any critical issues".

Frequently Asked Questions about auto-review

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

FAQPage Schema
How do I automatically review code changes after an AI agent turn?▼

The pi-auto-review extension queues a /skill:auto-review turn after an agent prompt changes files. The skill derives changed files with read-only git commands, dispatches parallel reviewer subagents, synthesizes findings, and optionally sends accepted fixes to one fixer subagent.

How do I configure reviewers, models, and skills for automated code review?▼

Use /auto-review config set commands to change reviewerAgent, reviewerSkills, reviewerProfiles, reviewConcurrency, and related keys. Reviewer profiles support per-profile agent, model, skills, and task prompts, merged by id across global and project config.

Can I run code review without auto-fixing the findings?▼

Yes, run /auto-review config set autoFix false to make the workflow review-only. The reviewers still report Critical, Warnings, and Suggestions, but no fixer subagent is dispatched and no files are edited.

Why was my auto review skipped after a commit?▼

The skill skips review when the worktree is clean with no committed range, when changes are only lockfiles or generated files, when the diff matches an already-reviewed pass, or when a committed range is a bookkeeping follow-up. It prints a single muted status line explaining the skip.

What are the limits of the parallel reviewer fanout?▼

Review concurrency is capped at 8, and a single subagent call supports at most 8 tasks, so larger reviewer lists are split into sequential batches. Reviewers are read-only; only the single fixer subagent may write files.

Related Skills

Explore More Software Engineering Skills→
timi-tytimi-ty

commit-agent-changes

Groups agent session changes into logical commits and opens a GitHub pull request.

Community
Advanced
human-centric-engineeringhuman-centric-engineering

pr-gates

Runs pre-merge validation, coverage, security, and code review gates on a branch and fixes findings until clean.

Official
Advanced
redpanda-dataredpanda-data
34

domain-model

Review domain models using product, engineering, and design perspectives.

Official
Advanced
Open-PawsOpen-Paws
3

code-review

Automate layered code review for pull requests in CI pipelines.

Official
Advanced