AgentSkillsCN

agent-skill-creator

创建、更新或审核可在不同 AI 代理间通用的代理技能。当用户提出撰写新的 SKILL.md 文件、将现有说明转换为代理技能格式、打包脚本/参考文献/资产，或确保与代理技能规范的兼容性与有效性时，此技能便能派上用场。

SKILL.md
--- frontmatter
name: agent-skill-creator
description: Create, update, or review Agent Skills that are portable across AI agents. Use when a user asks to author a new SKILL.md, convert existing instructions into the Agent Skills format, package scripts/references/assets, or ensure compatibility and validation against the Agent Skills specification.

Agent Skill Creator

Overview

Design Agent Skills that are spec-compliant, portable, and clear about when to activate. Prioritize concise triggers, progressive disclosure, and compatibility with both filesystem-based and tool-based agents.

Workflow

0. Confirm target agent constraints

  • •Ask which agent(s) the skill must work with and whether they are filesystem-based or tool-based.
  • •Ask about constraints: shell access, network access, allowed tools, and whether scripts can execute.
  • •If constraints are unknown, assume conservative defaults: no shell access, no network, no external dependencies. Keep scripts optional and provide manual steps.
  • •Include the compatibility frontmatter field only when a real constraint exists.

1. Gather concrete usage examples

  • •Ask for 3-5 example user requests that should trigger the skill.
  • •Ask for a couple of out-of-scope examples to set boundaries.
  • •Convert these into the skill's description keywords and "Examples" section.

2. Define scope and resources

  • •Decide which content belongs in SKILL.md vs references/ vs assets/.
  • •Use references/ for long docs, schemas, or policies.
  • •Use assets/ for templates or files that should be copied into outputs.
  • •Use scripts/ only if the target agent can execute them; otherwise provide manual alternatives.
  • •Keep the skill folder minimal and avoid extra documentation.

3. Author SKILL.md

  • •Follow the Agent Skills spec (summary in references/agentskills-spec.md).
  • •Frontmatter must include name and description with correct constraints.
  • •Keep the body imperative and task-focused: steps, edge cases, examples, and outputs.
  • •Keep SKILL.md under 500 lines and move details into references/.
  • •When referencing files, use relative paths from the skill root and keep references one level deep.

4. Add resources

  • •Add reference files with focused scope and clear titles.
  • •If scripts are included, document dependencies and a fallback path.
  • •Link to any references from SKILL.md using relative paths.

5. Validate

  • •If available, run skills-ref validate <skill-dir>.
  • •Otherwise, manually verify:
    • •name matches the folder and follows naming rules.
    • •description is specific and under 1024 chars.
    • •Optional fields are used only when needed.
    • •File references are relative and one level deep.

Output requirements

  • •Provide a concise file tree of the created skill.
  • •Include the final SKILL.md content.
  • •Call out any assumptions about agent capabilities.
  • •Provide 2-3 example prompts that should trigger the skill.

Examples

Example 1: Input: "Create a skill that helps agents generate weekly status updates for engineering teams." Output: A spec-compliant skill folder with SKILL.md instructions, a references/ guide for update format, and clear triggering keywords.

Example 2: Input: "Convert our internal onboarding doc into a reusable skill for any agent." Output: A skill that summarizes the onboarding workflow in SKILL.md with detailed policies in references/ and optional templates in assets/.

Response checklist

  • •Confirmed target agent constraints
  • •Defined scope and examples
  • •SKILL.md frontmatter valid and descriptive
  • •Progressive disclosure applied (references/assets/scripts)
  • •Validation completed or manual checks noted