
Writing Skills
If you're building skills for Claude, this is your meta-guide. It treats skill writing as test-driven development for documentation: run a baseline test with a subagent to watch it fail without the skill, write the skill to fix that specific behavior, then verify compliance. The core insight is about Claude Search Optimization, where descriptions must contain triggering conditions, not workflow summaries, because testing showed Claude will follow the description and skip reading the actual skill content. Includes concrete structure templates, naming conventions, and a detailed breakdown of what makes skills discoverable. Assumes you already understand TDD fundamentals from the aegis skill. Most useful when you're actually creating or debugging skills that Claude isn't finding or following correctly.
- GitHub stars
- 492
- last updated
- Jun 11, 2026
Install
$ npx -y skills add ganyuanran/aegis --skill writing-skills --agent claude-codeInstalls into .claude/skills of the current project.
Files
Writing Skills
Purpose
Create the smallest reusable skill that changes decisions or improves task
execution without constraining unrelated work. Skill changes use behavior
validation, not a code TDD route. This validation does not require loading aegis:test-driven-development.
Default Path
1. Establish Need And Ownership
- Name the reusable behavior gap, intended users/hosts, and non-goals.
- Check whether an existing skill, project rule, host adapter, or automated check already owns it. Prefer the smallest existing owner.
- For an existing skill, inspect its references, manifests or generated copies, discovery tests, and bundled resources before changing or retiring them.
- Create a skill only for reusable, non-obvious guidance. Keep one-off project conventions in the project owner and automate mechanical constraints when practical.
2. Record Behavior Before Editing
Use representative tasks, not questions that merely ask the agent to recite instructions.
Sponsors
Details
- Repository
- github.com/ganyuanran/aegis
- Categories
- Documentation, Skill Development
- License
- MIT
- First seen
- Jun 11, 2026
More from ganyuanran/aegis
All 22- Anti Entropy Governance
- Brainstorming
- Communicating Concisely
- Dispatching Parallel Agents
- Establishing Project Context
- Executing Plans
- Finishing A Development Branch
- First Principles Review
- Goal Framing
- Long Task Continuation
- Receiving Code Review
- Recording Architecture Decisions
- Requesting Code Review
- Subagent Driven Development
- Systematic Debugging
- Test Driven Development
- Update Aegis
- Using Aegis
- Using Git Worktrees
- Verification Before Completion
- Writing Plans