| name | start-project |
| description | Scaffold a new project to standard defaults and converge it — one command from idea to working skeleton |
Start Project
Use when the user wants to begin a new project or asks "where do I start". Produce a working skeleton, not a plan document. Give the user one explicit "start here" command at the end.
Procedure
- Ask the minimum: project name, one-sentence purpose, primary language (default Python). Do not interview beyond what scaffolding needs.
- Scaffold (Python default; adapt structure per the language rules in AGENTS.md for rust/go/etc.):
pyproject.toml only (no setup.py/setup.cfg/requirements.txt): hatchling, [project.scripts] entry, version 0.1.0.
- src layout:
src/<pkg>/__init__.py (with __version__ = "0.1.0"), __main__.py, cli.py (click, --version from package metadata).
tests/ with one real smoke test that runs via pytest.
README.md (purpose, install via pipx install ., usage), CHANGELOG.md (0.1.0 entry), LICENSE, .gitignore.
- Stack defaults: SQLite for authoritative state (derived state rebuildable); standard library first; logging module, type hints, Google docstrings.
- Initialize:
git init, initial commit (feat: initial project scaffold).
- Converge: run
crules <language> so the swarm skeleton, AGENTS.md, shims, and skills deploy immediately.
- Verify:
python3 -m venv .venv && .venv/bin/pip install -e . && .venv/bin/pytest must pass before declaring done.
- Hand off: tell the user the single next command to run, and note that the Bootstrapper interview will customize the swarm on first agent session.