FelipeOFF

PR Autopilot — Git skill for Claude Code

Git community

Claude Code skill that orchestrates the full Pull Request lifecycle — creation, multi-agent review, automated fixes, CI polling, and auto-merge.

How to install PR Autopilot

This entry records only its repository, not the path inside it, so there is no exact command to give. Open FelipeOFF/pr-autopilot and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What PR Autopilot does

Claude Code skill that orchestrates the full Pull Request lifecycle — creation, multi-agent review, automated fixes, CI polling, and auto-merge. Supports GitHub & GitLab.

Alternatives in Git

README

pr-autopilot

[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](./LICENSE) [![Claude Code Skill](https://img.shields.io/badge/Claude%20Code-Skill-7c3aed)](https://docs.claude.com/en/docs/claude-code) [![Platforms](https://img.shields.io/badge/platforms-GitHub%20%7C%20GitLab-blue)]()

🇧🇷 [Leia em Português](./README.pt-BR.md)

A [Claude Code skill](https://docs.claude.com/en/docs/claude-code/skills) that orchestrates the **full lifecycle of a Pull Request** with multiple coordinated subagents.

**Create → Review → Triage every comment → Respond → Re-review → Resolve conflicts & fix CI → Wait for CI → Merge.** Hands-off.

**Opt-in by default.** Every stage is off unless you ask for it. Bare `pr-autopilot` just opens the PR and stops. You switch on each stage with a flag (`--review`, `--resolve`, `--merge`) or turn them all on at once with `--auto`. `--cascade` is a separate opt-in: it ships a GitHub work item as a PR against the trunk. `--auto` does not turn it on.


What it does

`pr-autopilot` turns a long, manual PR ritual into a single command. It spawns a **Reviewer** subagent that audits your diff and an **Author** subagent that clears everything standing between the branch and a clean merge: every comment already on the PR (yours, your teammates', Copilot's, CodeRabbit's, Sonar's), merge conflicts, and failing CI. It loops until the PR is green, then merges.

**The Reviewer runs two parallel tracks** when `--review` or `--auto` is set: (A) a **code track** — loads [`thermo-nuclear-code-quality-review`](https://github.com/cursor/plugins/tree/main/cursor-team-kit/skills/thermo-nuclear-code-quality-review) via Skill tool for deep maintainability audit: code judo, abstraction quality, file size (<1k → >1k is a smell), spaghetti growth, structural simplification. This is the approval bar: do not approve merely because behavior works. Few high-conviction comments; no nit floods. (B) a **test track** — test-value assessment that evalu