Skip to main content
repo-ai

Turns ai-ready GitHub issues into reviewed PRs — one worktree per issue, two agent reviewers, and a human who merges.

Skills, labels, agent identity and statusline — asking before each
npx @rtorcato/repo-ai setup
Label-driven
Label-driven
All state lives in GitHub labels, never in a conversation — a missed tick, a crash, or a restart costs nothing.
One worktree per issue
One worktree per issue
Each ai-ready issue gets its own git worktree and branch, so agents run in parallel without touching your checkout.
Two agent reviewers
Two agent reviewers
A code reviewer and a security reviewer read every diff. Changes requested get up to two fix rounds, then a human.
You merge
You merge
The loop never merges an issue PR. It labels it merge-ready, assigns it to you, and says why on the PR when it holds back.

How it fits together

A label state machine, a stateless tick, and the skills that drive it. Start with the guide; doctor tells you what your repo is missing.

Read the guide →

Sibling projects

More from @rtorcato — same conventions, same release pipeline.