loopkit

Why we included this project

loopkit is a drop-in `.claude/` harness plus a library of 49 small skills, aimed at teams that want their coding agents to follow a repeatable workflow instead of improvising. The harness is just files on disk: settings, hooks, a verifier subagent, and an 8-line loop runner that enforce a Plan → Act → Verify cycle, with no daemon or runtime to babysit. The skills are plain markdown with YAML headers, so they load only when a task triggers them and nothing is Claude-specific; the same set carries across Claude Code, Cursor, Codex, and Gemini CLI. You can install the whole floor with one curl command or pull in a single skill via npx, and existing project files are left untouched by default. It is a lightweight, opinionated starting point for agent-driven development, and because the skills are just markdown, forking it to fit your own workflow is straightforward.