Repository navigation
Contributing
Thank you for wanting to make this toolkit better. This guide covers everything you need to submit a high-quality contribution.
- Types of Contributions
- Before You Start
- Adding a New Agent
- Adding a New Skill
- Adding Prompts
- Adding Examples
- Pull Request Process
- Quality Bar
- Code of Conduct
| Type | Welcome? | Notes |
|---|---|---|
| New agent | Yes | Must follow the agent template |
| New skill | Yes | Must follow the skill template |
| New prompts | Yes | Group by platform |
| Bug fix in existing agent | Yes | Describe the problem clearly |
| Example code | Yes | Must be minimal and self-contained |
| Documentation | Yes | Fix typos, add clarity |
| Refactor/restructure | Discuss first | Open an issue before touching structure |
- Search existing issues and PRs — your idea may already be in progress.
- Open an issue first for any new agent or significant change. Use the New Agent template.
- Keep scope small — one agent or one skill per PR. Bundled PRs are hard to review.
agents/android/
agents/ios/
agents/flutter/
agents/react-native/
agents/unity/
agents/unreal/
agents/cross-platform/
Use kebab-case. Be specific:
agents/android/memory-leak-detector/ # Good
agents/android/helper/ # Too vague
Copy templates/agent-template.md and fill in every section. Empty sections are grounds for rejection.
Required sections:
-
## Purpose— one sentence -
## Input Format— exact schema with field descriptions -
## Output Format— exact schema with field descriptions -
## System Prompt— the full prompt to paste into an LLM -
## Example— a real, runnable input/output pair -
## Notes— edge cases, limitations, version requirements
Add a matching code file in examples/<platform>/ that the agent can operate on.
A skill is a reusable prompt module — shorter than a full agent and composable with others.
- Pick the right platform directory under
skills/ - Copy templates/skill-template.md
- Fill in: purpose, when to use, the skill prompt, and one example
Prompts live in prompts/<platform>/. Each file should:
- Be a standalone
.mdfile with a descriptive filename - Start with a one-line description comment
- Include the raw prompt text and one example of expected output
Example files live in examples/<platform>/. Rules:
- Must compile or be syntactically valid
- Must be self-contained (no external dependencies not explained)
- Must have intentional issues or characteristics the paired agent can find
- Include a header comment pointing to the paired agent
- Fork the repo and create a branch:
feat/android-memory-leak-agent - Make your changes following the templates
- Update the README agent table if you added a new agent
- Submit the PR using the PR template
- A maintainer will review within 5 business days
PR checklist:
- Followed the correct template
- All sections in
agent.mdare filled in - Example input/output is real (not placeholder text)
- No sensitive data (API keys, real crash logs from production, PII)
- README table updated (for new agents)
Agents are rejected if they:
- Have placeholder examples (
<YOUR CODE HERE>with no explanation) - Duplicate an existing agent without meaningfully improving it
- Lack an output format definition
- Produce output that requires significant human cleanup to be useful
Agents are accepted if they:
- Solve a problem that actually slows developers down
- Have a concrete, testable example
- Produce structured, parseable output
- Work across common versions of the target platform
By contributing, you agree to follow the Code of Conduct. Be kind, be specific, be constructive.
Getting Started
Documentation
- All Documentation
- Installation
- Figma MCP and Design Handoff
- Figma Spec Tools
- Core Loops
- Mobile Engineering Loop
- Mobile Memory
- Performance Loop
Reference
- Agents
- Platform Guide
- Skills and Prompts
- Android Device QA Setup
- Mobile App Design Skill
- Mobile Harness
- Enterprise Mobile Harness
- Workflows
- Reference
Community
Repository github.com/salmanashraf/mobile-ai-agents
Version: v1.0.38 · MIT License