AgentSkillsCN

Skill Maker

技能制造者

SKILL.md

Skill Maker - Create or Update a Skill in This Library

Purpose

Create a new skill (or update an existing one) that provides a repeatable, step-by-step execution guide in this repository.

When to Use This Skill

Use when:

  • •No existing skill fits the task or workflow
  • •A repeatable process needs a standard playbook
  • •You need to update a skill after new learnings

Do NOT use when:

  • •Writing policies (use .cursor/rules/)
  • •Writing one-off notes or internal memos

Required Inputs

  1. •Skill name: Kebab-case (e.g., prd-creator, api-express)
  2. •Purpose: 1-2 sentences describing the task
  3. •Target scenarios: 3-6 concrete use cases
  4. •Required inputs: What the user must provide
  5. •Outputs: Files or artifacts produced
  6. •Related policies: Which .cursor/rules/ apply (if any)

Defaults:

  • •Skill location: skills/{skill-name}/
  • •Main file: skills/{skill-name}/skill.md
  • •Details files in skills/{skill-name}/details/

Steps

1) Confirm scope and reuse

  • •Check skills/INDEX.md and existing skills for overlap.
  • •If a skill is close, update it instead of creating a new one.

2) Create the skill structure

Create the directory and details files:

bash
mkdir -p skills/{skill-name}/details

Required files:

  • •skills/{skill-name}/skill.md
  • •skills/{skill-name}/details/README.md
  • •skills/{skill-name}/details/examples.md
  • •skills/{skill-name}/details/checklist.md
  • •skills/{skill-name}/details/anti-patterns.md

3) Write skill.md

Use this structure:

  • •Purpose
  • •When to Use / Do NOT Use
  • •Required Inputs
  • •Steps (clear, numbered)
  • •Expected Outputs
  • •Validation
  • •Related Skills
  • •See Also (details files)

Keep it concise and execution-focused.

4) Write details files

  • •README: guidance, patterns, and edge cases
  • •examples: 2-5 practical examples
  • •checklist: definition of done
  • •anti-patterns: common mistakes and fixes

5) Link to policies

  • •Reference relevant .cursor/rules/ in the skill or details.
  • •Keep policies in .cursor/rules/; do not duplicate them in skills.

6) Update model skill indexes (required)

After creating or updating a skill, update all model indexes:

  • •Codex: skills/INDEX.md
  • •Claude: skills/INDEX-CLAUDE.md
  • •Gemini: skills/INDEX-GEMINI.md

Also update:

  • •skills/README.md (new skill entry)
  • •.cursor/rules/120-skills-index.md (Codex skill picker)

7) Validate

  • •Paths and filenames are correct.
  • •No secrets or credentials are stored.
  • •Steps are clear and testable.
  • •Indexes and README include the new skill.

Expected Outputs

  • •New or updated skills/{skill-name}/ directory
  • •Updated skill indexes for Codex, Claude, and Gemini
  • •Updated skills/README.md and .cursor/rules/120-skills-index.md

Validation

  1. •Skill is discoverable from all index files.
  2. •Steps are actionable without extra clarification.
  3. •Details files are present and consistent.

Related Skills

  • •skills/00-template/skill.md - Template for skill structure

See Also

  • •Details: ./details/README.md
  • •Examples: ./details/examples.md
  • •Checklist: ./details/checklist.md
  • •Anti-Patterns: ./details/anti-patterns.md