Requesting a New Skill
We use a curated contribution model: community members request new skills via GitHub Issues, and maintainers build them.
Process
- Open an Issue using the New Skill Request template
- Fill in the details: repository URL, documentation paths, when-to-use triggers
- A maintainer reviews the request and clones the source repo to understand the tool
- The skill is built based on actual repository content (no assumptions)
- A PR is opened for review, and the skill is added to the installer
What Makes a Good Skill Request
- Clear source repository — the Git URL and branch
- Identified documentation — which files in the repo should be fetched
- Use-case description — when should the AI assistant activate this skill?
- Example interactions — what questions would a user ask?
Skill Structure
Each skill consists of:
skills/<name>/
config.sh # Repository URL, branch, doc paths
SKILL.md # Agent skill definition (YAML frontmatter + Markdown)
references/
REFERENCE.md # Index of available documentation
rules/
<name>.mdc # Optional Cursor-specific rules
Validation Checklist
Before a skill is merged, maintainers verify:
- Source repository is accessible and public
- Documentation paths exist in the repo
SKILL.mdaccurately describes the tool based on actual repo content- “When to Use” triggers are specific and useful
config.shvalues are correct- Tests pass on RHEL and macOS