Thank you for your interest in contributing! This document provides guidelines for contributing to these Agent Skills.
You can contribute in any of these ways:
- Submit a pull request with improvements.
- Raise an issue on GitHub for bugs, gaps, or enhancement ideas.
- Reach out on the Zoom Developer Forum with feedback and improvement suggestions for these agent skills.
- Documentation errors or outdated information
- Missing use cases or scenarios
- Incorrect code examples
- Fix typos and clarify explanations
- Add missing code examples
- Update for new SDK versions
- New use cases
- New platform coverage
- New integration patterns
- Fork the repository
- Make your changes
- Submit a pull request with a clear description
- Open an issue first to discuss the proposed change
- Fork the repository
- Create a feature branch
- Follow the skill format guidelines below
- Submit a pull request
---
name: skill-name
description: |
Brief description (1-3 sentences).
Include when to use this skill.
---
# Skill Title
[Content following the template in PLAN.md]- Keep SKILL.md under 500 lines - Move details to
references/ - Max 3 directory levels -
skill/references/file.md - Include code examples - Real, working code developers can use
- Document gotchas - Common mistakes and limitations
- Link to official sources - Prefer Zoom documentation
- Keep canonical skill folder names aligned with skills/SKILL.md.
- Current canonical folders include:
general,rest-api,webhooks,websockets,meeting-sdk,video-sdk,zoom-apps-sdkrtms,team-chat,ui-toolkit,cobrowse-sdk,oauth,zoom-mcpcontact-center,virtual-agent,phone,rivet-sdk,probe-sdk
- Use real markdown links for local docs (for example:
text -> docs/example.md). - Do not use backticks for local doc references if you want them counted in relationship graphs.
- Use repository-relative paths; do not commit machine-local absolute paths (for example
/home/your-user/...). - Every new
.mdfile should be linked from at least one parent/index/SKILL.mdfile.
After adding or moving docs, regenerate the relationship graph:
node md-graph/build-graph.mjs . md-graph/data.json
cp md-graph/data.json /path/to/your/hosted/skills-relationship/data.jsonGraph generation intentionally excludes non-skill content directories such as raw-docs, md-graph, tools, hidden directories, and dependency/vendor folders.
You can use Claude (or other AI assistants) to help create or improve skills:
-
Research Phase
Research the official Zoom documentation for [topic]. Check the developer forum for common issues. Find working code examples. -
Drafting Phase
Create a skill following the SKILL.md template. Include practical code examples. Document known limitations and gotchas. -
Validation Phase
Cross-check all information with official Zoom docs. Verify code examples are syntactically correct. Ensure links are valid.
- Be specific: "Create a use case for RTMS audio streaming to S3" not "write about RTMS"
- Provide context: Share relevant existing skills as examples
- Iterate: Review drafts and ask for improvements
- Verify: Always cross-check AI-generated content with official sources
| Task | How Claude Helps |
|---|---|
| Research | Search docs, forums, GitHub for information |
| Drafting | Create initial skill content following templates |
| Code examples | Generate working code snippets |
| Cross-referencing | Check consistency across skills |
| Formatting | Ensure markdown is correct |
| Task | Why Human Review |
|---|---|
| Technical accuracy | AI may hallucinate APIs or features |
| Real-world gotchas | Comes from actual development experience |
| Business logic | Zoom-specific requirements and policies |
| Security practices | Must be verified against official guidance |
- Verify all claims with official documentation
- Include working, tested code examples
- Document known limitations prominently
- Link to official resources
- Keep examples simple and practical
- Include unverified information
- Speculate about undocumented behavior
- Copy proprietary code without permission
- Include outdated or deprecated APIs without noting it
- Over-engineer examples
- Be respectful and constructive
- Focus on improving the documentation
- Credit sources appropriately
- Follow Zoom's developer terms of service
- Open a GitHub issue for questions about contributing
- Check existing issues before creating new ones
- Join the Zoom Developer Forum for Zoom-specific questions, feedback, and improvement requests for these agent skills
By contributing, you agree that your contributions will be licensed under the MIT License.