AgentSkillsCN

creating-agent-skills

创建并审核符合开放 Agent Skills 规范与最佳实践的 Agent 技能。当用户请求创建新技能、编写技能说明、生成 SKILL.md 文件、审查现有技能以确保合规性，或提升技能质量时，均可使用此功能。内容涵盖技能结构、命名规范、渐进式披露、撰写高效描述、设计操作与标准，以及验证流程。

SKILL.md
--- frontmatter
name: creating-agent-skills
description: 'Create and review Agent Skills that follow the open Agent Skills specification and best practices. Use when a user asks to create a new skill, build skill instructions, generate a SKILL.md file, review an existing skill for compliance, or improve skill quality. Covers skill structure, naming conventions, progressive disclosure, writing effective descriptions, creating actions and standards, and validation.'
license: MIT
allowed-tools: Bash

Creating Agent Skills

Overview

This skill provides capabilities for creating and reviewing Agent Skills that comply with the Agent Skills specification and follow authoring best practices.

Agent Skills are a lightweight, open format for extending AI agent capabilities with specialized knowledge and workflows. A skill is a folder containing a SKILL.md file with metadata and instructions, optionally bundled with scripts, references, and assets.

Capabilities

CapabilityActionDescription
Createactions/create.mdGenerate a new, specification-compliant Agent Skill
Reviewactions/review.mdAnalyse an existing skill for compliance and improvements

Standards

This skill bundles the following standards in standards/:

StandardFileDescription
Specificationspecification.mdCore Agent Skills format specification
Namingnaming.mdNaming conventions for skills and files
Descriptionsdescriptions.mdWriting effective skill descriptions
Structurestructure.mdDirectory structure and progressive disclosure
Instructionsinstructions.mdWriting effective skill instructions
Checklistchecklist.mdConsolidated compliance and quality checklist

Principles

1. Concise is Key

The context window is a shared resource. Only add context the agent doesn't already have. Challenge each piece of information: "Does the agent really need this explanation?"

2. Progressive Disclosure

Structure skills for efficient context use:

  • •Metadata (~100 tokens): name and description loaded at startup
  • •Instructions (<5000 tokens recommended): Full SKILL.md body loaded when activated
  • •Resources (as needed): Additional files loaded only when required

3. Set Appropriate Degrees of Freedom

Match specificity to task fragility:

  • •High freedom: Text-based instructions for flexible tasks
  • •Medium freedom: Pseudocode or parameterised scripts
  • •Low freedom: Exact scripts for fragile, error-prone operations

4. Test with Real Usage

Build evaluations before extensive documentation. Test with representative tasks. Iterate based on observed agent behavior, not assumptions.

Usage

  1. •Load this skill manifest
  2. •Identify the required capability (create or review)
  3. •Load the bundled standards from standards/
  4. •Execute the action following actions/<capability>.md

References