Runboard

See your technical-leadership maturity scorecard in five minutes — local-first, no signup, nothing leaves your machine.

Runboard is a CLI that runs the Runboard framework: a nine-dimension maturity instrument (Build / Run / Plan × Team / Tools / Techniques), scored 1–5 with trajectory indicators. It renders your scorecard as a heatmap, tells you your single biggest constraint, and tracks whether things are getting better over time.

Quickstart

npx runboard@latest init      # scaffold .runboard/ in your repo
npx runboard@latest assess    # guided 9-dimension self-assessment
npx runboard@latest board     # your heatmap + average + biggest constraint

Then, once you have data:

npx runboard@latest roadmap       # Now / Next / Later plan from your constraint
npx runboard@latest board --html  # a shareable, self-contained board.html
npx runboard@latest pulse         # (after a 2nd assessment) what moved since last time
npx runboard@latest status        # one-screen current state

Local-first, no phone-home

Every command runs entirely on your machine. No account, no network calls, no telemetry. Your data lives in .runboard/ in your own repo — commit it, and its git history becomes your trajectory record.

Commands

| Command | What it does | |---------|--------------| | runboard init | Scaffold .runboard/ (config, rubric, assessments). Idempotent. | | runboard assess | Record a 9-dimension assessment (interactive, or --set for agents). | | runboard board [--html] | Render the heatmap; --html writes a shareable file. | | runboard pulse | Compare the two latest assessments; flag stuck dimensions. | | runboard roadmap | Now/Next/Later plan from your binding constraint. | | runboard report --type <t> | Render a board-ready report from a template. | | runboard status | One-screen current state. | | runboard skills install [--target <dir>] | Copy the bundled SKILL.md skills into an agent's skills directory. |

Use it from your AI assistant

Portable skills/ (SKILL.md) and an MCP server let Claude, Cursor, Codex, Copilot, Gemini, and other agents run the assessment conversation and persist results through the same core — the numbers are always computed by the tool, never by the model.

The MCP server ships in this same package and runs locally over stdio (no network, no account). Add it to your client with one zero-install command:

{
  "mcpServers": {
    "runboard": { "command": "npx", "args": ["-y", "runboard", "mcp"] }
  }
}

See docs/mcp.md for Claude Desktop, Cursor, and VS Code setup.

Installing the skills

SKILL.md is a cross-agent open format, so "installing" a skill just means copying its folder to wherever your agent looks for skills. skills install does that from the copy already bundled in this package — no download, no network:

npx runboard@latest skills install                 # auto-detect (Claude Code → .claude/skills/)
npx runboard@latest skills install --target .cursor/skills   # any other agent
npx runboard@latest skills install --dry-run       # preview; writes nothing
npx runboard@latest skills install --force         # overwrite existing copies (e.g. after upgrade)

Without --target it auto-detects a supported agent (v1: Claude Code, via a .claude/ directory). For any other agent, point --target at its skills directory. Existing skills are never overwritten without --force. Restart your agent afterward so it discovers them.

Contributing

See CONTRIBUTING.md. This is a public project; work starts from GitHub issues.

License

MIT