Turbo
A dev process built from composable skills for Claude Code and Codex: plan, implement, review, test, commit and PR
Install
Walk me through the Turbo setup. Read SETUP.md from the tobihagemann/turbo repo and follow the guide for your edition.Recommended: the agent clones the repo, copies skills and configures the environment.
This is third-party code. Review the repository files before installing.
What it does
Turbo builds a working loop out of plain markdown skills. /turboplan gauges task complexity and routes it either straight to implementation or through a drafted and refined plan. /implement and /implement-plan make the changes, and /finalize runs formatting, lint, tests, code review, smoke tests, commit and PR creation. Separate /audit and /onboard pipelines check project health and help you ramp up on a new codebase. Other skills cover bug investigation, threat models, test plans, dead code detection, PR comment replies and cross-review via Codex or Claude.
Who it is for. For experienced developers who want the agent to follow a full process from plan to PR with quality checks.
Good fit when
- You want a repeatable plan, implement, verify and PR loop
- You need a project health audit or to understand unfamiliar code
- You want changes reviewed by a second model
Not a fit when
- Your project lacks the tests and linters the loop relies on
- You only want a couple of standalone skills: many Turbo skills depend on each other
Example request
/turboplan add CSV export for reportsLimitations
The author recommends top-tier Claude Code or Codex plans because pipelines use a lot of context. Cross-review needs access to the other agent, and /consult-oracle drives ChatGPT in a browser and needs a subscription. Requires GitHub CLI. The Codex edition is still experimental.
How to disable. Delete the Turbo skill folders from ~/.claude/skills, the ~/.turbo directory and the rules added to CLAUDE.md during setup.
Security check
- Skills write code, commit and open PRs
- They run the project's tests, linters and app
README in short
The README describes the three-step core loop and two extra pipelines. It explains the principles: a skill as a ready prompt, nested skills, swappable parts and no custom runtime. It details the planning and finalize pipelines plus project and external service requirements. The agent performs installation from SETUP.md. MIT licensed.
FAQ
Can I swap skills for my own?
Yes, each skill owns one concern and talks to others via git, PR state and files. Commit rules or code style, for instance, can be replaced with your team's.
How do I update?
Run /update-turbo in Claude Code or $update-turbo in Codex. It builds a changelog and handles skills you customized.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything
Up-to-date, version-specific library docs and code examples in your agent's context, via MCP or a CLI plus skill