Audit, optimize, or create agent skills using Anthropic best practices. Use when reviewing existing skills for quality, writing new SKILL.md files, or finetuning skill descriptions, structure, and progressive disclosure patterns.
Evaluate and improve agent skill files using principles from Anthropic's skill authoring guide. Use this skill to audit existing .github/skills/*/SKILL.md files, create new skills, or refine skill metadata for better discovery and token efficiency.
Run through each item when reviewing a skill. Flag violations explicitly.
name: ≤64 chars, lowercase + hyphens only, no reserved words (, ).anthropicclaudedescription: non-empty, ≤1024 chars, written in third person.Match specificity to task fragility:
| Freedom | When | Style |
|---|---|---|
| High | Multiple valid approaches, context-dependent | Text instructions, heuristics |
| Medium | Preferred pattern exists, some variation OK | Pseudocode, parameterized templates |
| Low | Fragile/error-prone, consistency critical | Exact scripts, no modification allowed |
a.md → b.md → c.md chains).form_validation_rules.md, not doc2.md).ServerName:tool_name.Complete the task manually with the agent first. Note what context you repeatedly provide — that's the skill content.
---