Skip to content

Agents.md

Latest commit

 

History

History
232 lines (157 loc) · 4.31 KB

Agents.md

File metadata and controls

232 lines (157 loc) · 4.31 KB

AI Agents for WhatsApp CLI

This project includes custom AI agents for automation and development assistance.

Available Agents

TUI WhatsApp AI Developer

Location: .github/agents/tui-whatsapp-ai-dev.agent.md

A specialized agent for WhatsApp CLI development tasks.

Capabilities

  • WhatsApp Integration

    • whatsapp-web.js API usage
    • Puppeteer configuration for headless Chrome
    • QR code authentication flow
    • Session persistence management

  • CLI Development

    • readline-based menu systems
    • Terminal UI with chalk colors
    • Input validation and error handling
    • Settings management

  • AI Provider Integration

    • OpenRouter API integration
    • OpenAI GPT models
    • Google Gemini
    • API key management
    • Temperature and token configuration

  • Message Handling

    • Sending/receiving messages
    • Chat history management
    • Media message handling
    • Group chat operations

Usage Examples

# Using with Claude Code
"Use the tui-whatsapp-ai-dev agent to add voice message support"

# Direct agent invocation
/agent tui-whatsapp-ai-dev "Implement message search functionality"

Creating Custom Agents

Agent Template

# Agent Name

Brief description of what this agent does.

## Context

The agent should understand:

- Project structure
- Key dependencies
- Relevant code patterns

## Capabilities

- Capability 1
- Capability 2

## Instructions

When invoked, the agent should:

1. Read relevant files
2. Analyze the request
3. Propose solution
4. Implement changes
5. Verify with tests

Example: Message Automation Agent

# Message Automation Agent

Automates repetitive messaging tasks while respecting WhatsApp's rate limits.

## Context

- WhatsApp Web API for sending messages
- Rate limiting to prevent account bans
- Message queue management
- Scheduled message support

## Capabilities

- Bulk message sending with delays
- Scheduled message delivery
- Message templates
- Auto-reply rules

## Instructions

1. Validate recipient list
2. Check rate limit compliance
3. Queue messages with appropriate delays
4. Handle delivery failures
5. Log all activity

Agent Configuration

Agent Metadata

name: "tui-whatsapp-ai-dev"
version: "1.0.0"
description: "WhatsApp CLI development specialist"
author: "involvex"

Agent Permissions

Agents may request:

  • File read/write access
  • Package installation
  • Build/compile operations
  • Git operations

Security Considerations

Agent Sandboxing

  • Agents run in isolated environments
  • File access is scoped to project directory
  • External API calls require confirmation

Sensitive Data

Agents should NOT:

  • Store API keys in code
  • Commit credentials
  • Share session data
  • Expose message content

Best Practices

1. Clear Instructions

Provide specific, actionable requests:

# Good
"Add a new menu option for exporting chat history to JSON"

# Vague
"Add export feature"

2. Context Awareness

Include relevant context:

"Add message search to the main menu (option 9) that searches through loaded chats"

3. Verification

Ask agents to verify changes:

"Add the feature and run the build to verify it works"

Agent Development

Adding New Agents

  1. Create agent file in .github/agents/
  2. Define capabilities and context
  3. Test with various requests
  4. Document usage examples

Agent Testing

# Test agent behavior
"Use the tui-whatsapp-ai-dev agent to explain the message flow"

Troubleshooting

Agent Not Responding

  • Check agent file syntax
  • Verify agent is loaded
  • Review Claude Code logs

Agent Misunderstanding

  • Provide more context
  • Break down complex requests
  • Reference specific files

Future Agents

Planned Agents

  • Message Automation Agent - Bulk operations with rate limiting
  • Analytics Agent - Chat statistics and insights
  • Plugin Agent - Extensible plugin system
  • Testing Agent - Automated test generation

Contributing

To contribute a new agent:

  1. Create .github/agents/your-agent.agent.md
  2. Document capabilities and usage
  3. Add examples to this file
  4. Submit PR

Resources