| name | starting-a-project |
| description | Use when bootstrapping a new project repository for Kilroy Attractor from a clean directory using existing spec, DoD, graph, and run config artifacts. |
Starting A Project
Use this skill to prepare a clean repo for kilroy attractor validate/run. This skill does not create requirements artifacts; it only wires and verifies them.
Required Inputs
- Target repo path (clean directory)
- Existing project artifacts: spec (
*.md), DoD (*.md), graph (*.dot), run config (*.yaml or *.json)
- Path to the Kilroy source repo (for canonical skill content and helper scripts)
Workflow
- Create the target directory and initialize git.
- Add canonical skill mounts:
skills -> <kilroy-repo>/skills
.claude/skills/<name> -> ../../skills/<name>
.agents/skills/<name> -> ../../.claude/skills/<name>
- Copy project artifacts into the new repo (commonly under
demo/): spec, DoD, graph, run config, and any graph-referenced companion files.
- Update run config for the new repo:
- Set
repo.path to the absolute target repo path.
- Ensure absolute helper paths (for example
modeldb path or scripts/start-cxdb.sh) still resolve on this machine.
- Create a base commit so Attractor can create run branches/worktrees.
- Verify readiness:
kilroy attractor validate --graph <graph.dot>
- Confirm required executables and env vars expected by config/setup exist.
Guardrails
- Do not invent or regenerate spec/DoD/run config content in this skill.
- If any required artifact is missing or empty, stop and report the gap.
- Keep all setup deterministic and local-first (no production run side effects).
Done Checklist
- Target directory is a git repo with at least one commit.
- Skill symlink layout resolves (
skills, .claude/skills, .agents/skills).
- Spec, DoD, graph, and run config are present and non-empty.
repo.path in run config points at the new repo.
- Graph validation returns no errors.