Installs into .claude/skills of the current project.
Are you the author of Onboard?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/ahmedbutt2015-onboard)
# /onboard
Use this skill from a repository root to generate an onboarding tour for an unfamiliar codebase.
## Current workflow
1. Run the deterministic scanner to detect repo shape and entry points.
2. Extract a verified spine from supported local language relationships.
3. Generate and validate a Mermaid diagram from the verified spine.
4. Cluster the remaining files into heuristic subsystems.
5. Build a synthesis prompt that only contains verified structure and short code signatures.
6. **When running as Claude Code skill**: Write prompt to file, ask Claude to synthesize JSON, then read JSON back to generate `ONBOARDING.md`.
7. Write `ONBOARDING.md` into the target repo root.
## Local command
```bash
npm run onboard -- .
```
Optional token-free preview first:
```bash
npm run map -- .
```
Optional prompt export:
```bash
npm run onboard -- . --prompt-out .onboard-prompt.txt
```
Optional synthesis input (for Claude Code skill flow):
```bash
npm run onboard -- . --synthesis-input .onboard-response.json
```
Optional external synthesis command:
```bash
npm run onboard -- . --synthesis-command "your-command-here"
```
## Product contract
- Repo brand: `spine`
- User-facing Claude Code command: `/onboard`
- Companion map-only command: `/map`
- Diagram rule: verified edges only, never guessed
- Output target: `ONBOARDING.md`
## Launch positioning
The clearest user story is:
1. `/map` shows the verified architecture preview
2. `/onboard` writes the full guide
3. `.claude/REPO_CONTEXT.md` keeps future Claude sessions grounded in the repo
