Using with repo-ai (optional)

repo-tooling doesn't need repo-ai. Everything it does (setup, doctor, fix, the presets and the agent-rule files) works the same with or without it. Nothing in repo-tooling installs repo-ai, prompts for it, or fails a check because it's missing.

@rtorcato/repo-ai is optional. It adds an ai-issue-loop: a label-driven pipeline that takes a GitHub issue marked ai-ready, implements it in its own git worktree, opens a PR, has two agents review it, and hands it to you to merge. Adopt it on a repo when you want agents working your issue queue. Leave it out otherwise.

Who does what

	repo-tooling	repo-ai
Purpose	Scaffold and audit a repo's tooling: TypeScript, lint, tests, git hooks, CI, releases, GitHub settings	Run agents against ai-ready issues and carry their PRs to review
Needed by the other?	No	Builds on the repo standard below
CLI	repo-tooling setup / doctor / fix	repo-ai loop … / doctor / fix
Claude Code skills	repo-tooling, npm-publish, dogfood	ai-workflow, ai-issue-loop, ai-issue, ai-loop-status
Config	.repo-tooling.json	.repo-ai.json

What repo-ai relies on from repo-tooling

repo-ai doesn't import repo-tooling. It assumes a repo already meets the standard repo-tooling sets up, and each piece of that has its own fix target:

• Squash merges, auto-merge, required status checks, and no required approving reviews (fix github-settings). The loop's reviewers approve with labels, so a required human approval would deadlock every PR it opens.
• A release environment with required reviewers (fix release-environment), so a merge never reaches npm without a person approving it.
• Worktree dependency links in .claude/settings.json (fix ai), so each agent's worktree builds without a fresh install.
• Nothing in .repo-tooling.json: repo-ai keeps its own settings in .repo-ai.json. Older repos that still have rules.aiLoop or rules.requiredSkills there can move them across with npx @rtorcato/repo-ai fix config.

You can meet the same standard by hand. repo-tooling is just the quickest way.

Adding repo-ai to a repo

npx @rtorcato/repo-tooling doctor   # the repo standard above: fix anything it flags
npx @rtorcato/repo-ai setup         # skills, labels, agent identity, statusline, then its own doctor

repo-ai setup asks before each step, and --yes runs them all.

After that, label an issue ai-ready and run /ai-workflow in Claude Code. repo-ai's guide to the loop covers the label state machine, the safety gates and the limits. Its command reference covers every repo-ai command.

Moving over from repo-tooling 3.x

The loop shipped inside repo-tooling until 4.0.0.

Was (repo-tooling 3.x)	Now (repo-ai)
repo-tooling loop <guard|env|tick|worktree add|cleanup|reap|comment|verdict>	repo-ai loop …, same subcommands and flags
repo-tooling fix claude-skills	repo-ai fix claude-skills
repo-tooling fix labels	repo-ai fix labels
repo-tooling fix ai-loop-identity	repo-ai fix ai-loop-identity
The AI loop labels, AI loop agent, Claude skills and Required skills doctor checks	repo-ai doctor
The ai-issue-loop, ai-workflow, ai-issue and ai-loop-status skills	Shipped by repo-ai

repo-tooling loop … now only prints a pointer to repo-ai and exits 1.

Reinstall the skills once with --force-skills:

npx @rtorcato/repo-ai fix claude-skills --force-skills

Skills installed by repo-tooling carry its version stamp, so repo-ai can't tell them from a local edit and won't overwrite them without the flag.