AgentSkillsCN

authoring-skills

在创建、优化或评审技能时，务必使用此功能。当用户提及“创建技能”“新技能”“撰写技能”“优化技能”“技能描述”或“渐进式披露”时，请主动启用该功能。它以三级渐进式披露的方式，引导开发者设计符合规范的技能。

SKILL.md
--- frontmatter
name: authoring-skills
description: MUST BE USED when creating, improving, or reviewing Skills. Use PROACTIVELY when user mentions "create a skill", "new skill", "write a skill", "improve skill", "skill description", or "progressive disclosure". Guides through specification-compliant skill design with three-level progressive disclosure.
metadata:
  category: meta
  keywords: skill, create, authoring, develop, progressive-disclosure, compliance

Skill Development

Guides effective Skill creation following the Universal Skills Specification.

Overview

Skills enable agents to work autonomously without human intervention. This skill provides the methodology for creating specification-compliant skills using progressive disclosure, robust scripting, and agentic patterns.

When to Use

  • •User mentions "create a skill", "new skill", "write a skill", "improve skill"
  • •Reviewing or auditing existing skills for compliance
  • •Designing skill architecture with proper agentic patterns

Workflow

<workflow> 1. **Analyze & Plan**: - Determine Degrees of Freedom (Fragile vs Creative) using `references/agentic-patterns.md`. - Define Security Scope using `references/security-best-practices.md`. - Create `evaluation_plan.md` from `assets/templates/evaluation_plan.md`.
  1. •

    Draft Content:

    • •Use assets/templates/skill_structure.md for SKILL.md.
    • •Use assets/templates/robust_script.py for any Python scripts.
    • •Write frontmatter using references/description-templates.md (CRITICAL).
  2. •

    Verify & Refine:

    • •Validate against references/compliance-checklist.md.
    • •Ensure all paths are relative to project root.
    • •Verify that instructions use imperative, third-person language (no "you"). </workflow>

Critical Rules

<rules> - **Frontmatter**: MUST use directive language (MUST BE USED, Use PROACTIVELY) in `description`. - **Language**: NEVER use second person ("You should"). Use imperative ("Do X"). - **Structure**: Keep `SKILL.md` body under 500 lines. Move details to `references/`. - **Pathing**: ALWAYS use project-root relative paths. - **Verification**: workflows MUST include self-correction loops. </rules>

Resources

  • •Core Principles: references/core-principles.md (Philosophy, Directory Structure, Detailed Workflow)
  • •Templates:
    • •Description: references/description-templates.md
    • •Structure: assets/templates/skill_structure.md
    • •Script: assets/templates/robust_script.py
    • •Evaluation: assets/templates/evaluation_plan.md
  • •Guides:
    • •references/compliance-checklist.md
    • •references/agentic-patterns.md
    • •references/security-best-practices.md
    • •references/progressive-disclosure.md