Skip to content
Jackdaws

Claude Skills, treated like a real product

Design, build, grade, and ship Claude Skills through hard gates.

Jackdaws is an opinionated production line for Claude Skills. Numbered rules and hard gates replace subjective review, so a skill library earns the same rigor as production software.

  1. 01

    Design

    Turn a one-paragraph brief into a typed, validated design plan.

  2. 02

    Build

    Drive the plan through a six-stage lifecycle with per-stage status.

  3. 03

    Grade

    Score the result against numbered rules — Q001–Q010, not opinions.

  4. 04

    Ship

    One GO / NO-GO gate chains preflight, grader, and ecosystem checks.

    ship_checkGO/ NO-GO
66skills in the toolkit
72read-only verifier tools
51numbered rule families
13composable workflows

The difference

Numbered rules, not vibes

Every check is a named, mechanical rule with a stable id. Quality grading, upload validation, doc-drift, context engineering, and security posture each carry a numbered family. A finding points at a specific rule, not a reviewer’s mood.

Q001–Q010P001–P017DD001–DD006CE001–CE007LS001–LS010MS001–MS020

The terminal gate

ship_check chains preflight, the quality grader, and the ecosystem audit into a single binary verdict. The result is in the output, not a subjective sign-off: a skill is either GO or NO-GO.

GO/ NO-GO

Beyond skills

The toolkit eats its own dog food

Scaffold MCP servers

Generate and audit Model Context Protocol servers against MS001–MS020.

Audit whole repos

Doc drift, context engineering, LLM security, and review readiness.

Account cost

Report the token and dollar cost of any workflow run, per model tier.

Lint the copy

The copy deck behind this site passes a numbered voice guide, V001–V010.

Ways to use it

Hosted server, or done with you

Connect your client

claude.ai, Claude Code, Google Antigravity, or your own harness — one credential, and per-client steps for each.

See what is behind it

91 tools over one authenticated MCP connection — the verifiers plus generative agents that design, bridge, and eval skills.

Bring in Hazeley

Assess, build, or standardize your skill library with the toolkit's gates as the acceptance criteria.

Built for teams standardizing on Skills

Claude Code power users, consultants, and teams who want their skill library treated like a real software product — versioned, graded, and shipped through gates.