agenture-cli

Bootstrap your repo with a shared memory system, agent rules, and skill install playbooks — compatible with Claude Code, GitHub Copilot, Cursor, and OpenAI Codex.

What it does

Phase 1 — init: copies a fixed set of files into your project (no prompts, no network):

| What | Where | |---|---| | Shared memory system | .memory/ + AGENT_MEMORY_RULES.md | | Agent hub | .agents/ (skills land here for most tools) | | Universal agent instructions | AGENTS.md | | Skills install playbook | AGENT_SKILLS_INSTALL.md | | Claude Code | CLAUDE.md + .claude/commands/setup-agenture.md | | GitHub Copilot | .github/copilot-instructions.md | | Cursor | .cursor/rules/agent-memory.mdc |

Phase 2 — skills: your agent reads the repo, searches skills.sh, and installs the skills that match your stack via npx skills.

Requirements

• Node.js 20+

Usage

On any existing project

cd /path/to/your/project
npx agenture-cli@latest init

Existing files are left untouched. Use --force to replace them:

npx agenture-cli@latest init --force

From a clone of this repo

pnpm install
pnpm run build
node dist/cli.js init /path/to/your/project

Check your environment

npx agenture-cli@latest doctor

Prints your Node/npx versions and skills CLI pointers.

Phase 2: install skills

After init, the CLI prints a prompt to paste into your agent. It will:

1. Scan your stack (package.json, lockfiles, frameworks, etc.)
2. Search skills.sh with npx skills find
3. Install the relevant skills with npx skills add … -a <agent>

Claude Code — run the slash command:

/setup-agenture

Copilot / Cursor / Codex / other agents — paste the prompt printed by init, or tell your agent:

Read AGENT_SKILLS_INSTALL.md at the repo root and execute phase 2 — real npx skills commands, use the right -a for your agent (github-copilot, cursor, codex…).

Skills are installed into .agents/skills/ (project-scoped) by default. Commit that folder to share them with your team.

Memory system

Every agent that reads the bootstrapped files follows the same memory discipline defined in AGENT_MEMORY_RULES.md:

• Decisions, preferences, conventions, and patterns go into .memory/{category}/YYYY-MM-DD-slug.md
• The agent updates .memory/INDEX.md after each write
• No keyword needed — any lasting project signal triggers a write automatically

Categories: decision/, preference/, context/, pattern/

Telemetry

The skills CLI (npx skills) collects usage telemetry by default. Opt out:

export DISABLE_TELEMETRY=1
# or
export DO_NOT_TRACK=1

License

MIT