tobihagemann/turbo
A composable dev process for agentic coding harnesses, packaged as modular skills. Turbo has sibling editions for Claude Code and Codex.
What it solves
Turbo provides a standardized, modular dev process for agentic coding harnesses (like Claude Code and Codex), replacing vague prompting with a structured lifecycle. It solves the problem of AI agents skipping edge cases, missing project patterns, or producing inconsistent quality by wrapping the development loop into a set of composable "skills" (markdown-based prompts).
How it works
Turbo operates as a library of 70+ modular skills that integrate directly into an AI coding agent's native primitives. It organizes development into three primary pipelines:
- The Turboplan Pipeline: A universal entry point that analyzes task complexity to route work through either a "Direct mode" (for small changes) or a "Plan mode" (for which a formal plan file is drafted, refined, and reviewed).
- The Implementation Phase: Executes the plan or ad-hoc changes, incorporating code style guides and smoke testing for UI/UX changes.
- The Finalize Pipeline: A QA and commit loop that automatically runs tests, formats code, lints, cleans up documentation, updates changelogs, and handles the git commit/PR process.
Additionally, it includes out-of-loop pipelines for project-wide health audits and developer onboarding, and a /self-improve skill that extracts lessons from conversations to update project instructions and memory.
Who it’s for
Experienced developers who want to accelerate their workflow without sacrificing quality, as well as beginners who want to learn professional development patterns through AI-assisted coding.
Highlights
- Composable Design: Skills can be nested and composed into complex pipelines (e.g.,
/finalizewraps multiple other skills). - Self-Improving: Automatically routes session learnings back into project memory and skills to improve future performance.
- Edition-Specific: Tailored versions for Claude Code and Codex to leverage their specific toolsets and MCP servers.
- Puzzle Piece Philosophy: Every skill is a self-contained markdown file that can be swapped or replaced with team-specific conventions.
- Integrated QA: Built-in loops for formatting, linting, and testing before any code is committed.
관련
- 프로젝트
- 프로젝트
- Dispatch
- 프로젝트
- Dispatch