skill-authoring

Create or update project-specific skills in this repo. Use when asked to make new skills, modularize workflows, or build skill packs for this codebase.

$ 安裝

git clone https://github.com/majiayu000/claude-skill-registry /tmp/claude-skill-registry && cp -r /tmp/claude-skill-registry/skills/development/skill-authoring ~/.claude/skills/claude-skill-registry

// tip: Run this command in your terminal to install the skill


name: skill-authoring description: Create or update project-specific skills in this repo. Use when asked to make new skills, modularize workflows, or build skill packs for this codebase.

Skill Authoring

Overview

Create concise, modular skills stored under skills/ in this repo.

Workflow

  1. Define triggers and scope with concrete user examples.
  2. Choose a short skill name (lowercase, hyphen).
  3. Initialize with init_skill.py from the system skill-creator.
  4. Write SKILL.md in imperative form and keep it lean.
  5. Add scripts or references only when they save repeated work.
  6. Package with package_skill.py only when requested.
  7. Run the QA harness to generate matrices and results templates.

References

  • Skill template: references/skill_template.md
  • QA checklist: references/qa_checklist.md
  • QA matrix template: references/qa_matrix_template.md

QA Harness

  • Generate QA artifacts: python3 scripts/skill_qa_harness.py
  • Overwrite existing outputs: python3 scripts/skill_qa_harness.py --force

Local Conventions

  • Prefer small, single-purpose skills plus an orchestrator.
  • Put reusable commands in scripts; keep references one hop from SKILL.md.
  • Avoid extra docs (README, changelog, etc.).
  • Add honesty guardrails when a skill can produce overconfident outputs.
  • Add acceptance criteria sections so QA can be objective.
  • Use repo-relative paths in skill lists (AGENTS.md) for portability.