Install Michi

Prerequisites

• Claude Code installed and working
• A project with a git repository
• The Michi toolkit — clone or download the GitHub repo

Install the skills

Copy the skill directories from the toolkit into your project’s .claude/skills/:

# From your project root
cp -r <michi-path>/toolkit/skills/michi-bootstrap .claude/skills/
cp -r <michi-path>/toolkit/skills/michi-planning .claude/skills/
cp -r <michi-path>/toolkit/skills/michi-session .claude/skills/
cp -r <michi-path>/toolkit/skills/michi-debrief .claude/skills/
cp -r <michi-path>/toolkit/skills/michi-sustainability .claude/skills/
cp -r <michi-path>/toolkit/skills/michi-explore .claude/skills/
cp -r <michi-path>/toolkit/skills/michi-workshop .claude/skills/
cp -r <michi-path>/toolkit/skills/michi-scenario-test-builder .claude/skills/
cp -r <michi-path>/toolkit/skills/michi-pr-prep .claude/skills/

Each skill is self-contained — its references/ folder includes everything it needs. No external dependencies.

Alternate: install via agent-sync

If you use agent-sync to manage skills across machines and tools (Claude Code, Codex, Gemini), the published toolkit/skills-directory.yaml already references the Michi skills hosted on GitHub. Add an entry pointing at this repo to your source directory’s skills-directory.yaml, or copy the published file directly:

# from a checkout of csepulv/michi
cp toolkit/skills-directory.yaml ~/ai-config/personal/

# fetch and sync
agent-sync fetch
agent-sync sync

Released versions are listed in CHANGELOG.md. The published repo tracks main only — to pin to a specific release, replace tree/main/ with the corresponding public commit SHA in each source URL.

The skills at a glance

Skill	When	What it does
michi-bootstrap	Before first session	Survey project, assess gaps, scaffold Michi docs structure
michi-explore	Research, brainstorming	Structured investigation with Michi discipline
michi-planning	Before implementation	Explore codebase, surface assumptions, co-design scenarios, write plan
michi-session	During implementation	Rigid execution: implement, test, log decisions, verify, commit
michi-debrief	After implementation	Review decisions, promote learnings, curate scenarios, calibrate trust
michi-sustainability	At checkpoints	Scaled health checks — code quality, test quality, architectural alignment
michi-scenario-test-builder	During planning	Generate Kaner-style verification scenarios
michi-workshop	Everyday work	Lightweight Michi for bug fixes, small features, quick tasks
michi-pr-prep	Before PR review	Prepare a review guide for your PR

Verify it works

Start a Claude Code session in your project and invoke:

/michi-bootstrap

If the skill loads and starts surveying your project, you’re set. Next: bootstrap your project →