| name | example-skill |
| description | Example skill demonstrating the registry structure and format |
| author | Skills Registry Team |
| date | Tue Dec 10 2024 00:00:00 GMT+0000 (Coordinated Universal Time) |
example-skill - Research Notes
Experiment Overview
| Item | Details |
|---|---|
| Date | 2024-12-10 |
| Goal | Create an example skill to demonstrate the skills registry format |
| Environment | Any system with Claude Code installed |
| Status | Success |
Context
When setting up a skills registry, it helps to have a concrete example showing the expected structure and format. This skill serves as that reference.
Verified Workflow
- Copy the template folder:
cp -r templates/experiment-skill-template plugins/training/your-skill-name
- Rename the skills subdirectory:
mv plugins/training/your-skill-name/skills/TEMPLATE_NAME \
plugins/training/your-skill-name/skills/your-skill-name
- Update plugin.json with your skill's metadata:
{
"name": "your-skill-name",
"version": "1.0.0",
"description": "Specific triggers: (1) when X, (2) when Y",
"author": { "name": "Your Name" },
"skills": "./skills",
"repository": "https://github.com/smith-cop/Skills_Registry"
}
Fill in SKILL.md with your learnings, especially the Failed Attempts table.
Validate before committing:
python scripts/validate_plugins.py
Failed Attempts (Critical)
| Attempt | Why it Failed | Lesson Learned |
|---|---|---|
| Generic descriptions | Skills weren't discoverable by /advise | Always include specific trigger conditions |
| Omitting failed attempts | Teammates repeated same mistakes | Document failures even more than successes |
| Vague parameters | Results weren't reproducible | Include exact, copy-pasteable configs |
Final Parameters
The minimum required plugin.json structure:
{
"name": "skill-name",
"description": "Trigger conditions...",
"skills": "./skills"
}
Key Insights
- The Failed Attempts table is referenced more than any other section
- Specific trigger conditions in descriptions make /advise more useful
- Include environment details for reproducibility
- Keep skills focused - one experiment per skill
References
- Claude Code Documentation
- Instructions from the skills registry blog post