Agent Communication Debugger Skill

A comprehensive debugging skill for the A2A (Agent-to-Agent) communication system.

What This Skill Does

Provides systematic diagnosis and troubleshooting for:

• Agent status and health checks
• Message routing issues
• Transport layer connectivity
• Configuration validation
• Log analysis
• Common problem resolution

When Claude Uses This Skill

Claude will automatically load this skill when you ask about:

• "debug agent communication"
• "why isn't the coder agent responding"
• "check agent status"
• "orchestrator routing problems"
• "messages not being delivered"
• "agent health check"

Skill Structure

agent-debug/
├── SKILL.md               # Main skill instructions
├── README.md              # This file
├── common_issues.md       # Detailed troubleshooting guide
└── scripts/
    └── test_message.py    # Message delivery test script

Quick Start

Using the Skill

Simply ask Claude questions like:

Can you debug why my agents aren't communicating?

Check the status of the orchestrator agent

The coder agent isn't responding to messages, help me diagnose

Claude will:

1. Load the agent-debug skill automatically
2. Follow the diagnostic workflow
3. Run necessary checks
4. Provide specific solutions

Running the Test Script Manually

You can also run the test script directly:

python .claude/skills/agent-debug/scripts/test_message.py

This will:

• Send a test message to the orchestrator
• Check for responses
• Verify agent processes are running
• Provide a diagnostic summary

Features

Systematic Diagnostics

The skill provides step-by-step diagnostics:

1. Check agent status (running/stopped)
2. Inspect configurations (agent.json files)
3. Analyze logs for errors
4. Verify message transport
5. Test message delivery
6. Diagnose routing issues
7. Check environment variables
8. Restart agents if needed

Common Issues Guide

The common_issues.md file covers:

• Messages not being delivered
• Routing to wrong agent
• Agent not generating responses
• Duplicate message processing
• Transport connectivity problems
• Orchestrator not starting

Quick Diagnostic Checklist

A handy checklist to run through:

• All required agents running
• WebSocket server active (if used)
• Agent configs valid
• Agents discovered by orchestrator
• API keys set
• Recent log activity
• No Python exceptions
• Test message succeeds
• Routing selects correct agent

Example Usage

Example 1: Quick Health Check

You: Check if all agents are running properly

Claude will:

1. Run ps aux to check processes
2. Verify orchestrator, coder, tester agents
3. Check logs for recent activity
4. Report status and any issues

Example 2: Routing Problem

You: Messages are going to dashboard-agent instead of coder-agent

Claude will:

1. Check orchestrator's routing decisions in logs
2. Verify coder-agent is discovered
3. Check routing keywords and LLM prompts
4. Suggest fixes (update keywords, restart with API key)

Example 3: No Response

You: Orchestrator receives my message but nothing happens

Claude will:

1. Check orchestrator logs for processing
2. Verify routing decision
3. Check target agent is running
4. Check target agent logs for errors
5. Identify missing dependencies or API keys

Integration with Agent System

This skill is designed specifically for the agent system in:

a2a_communicating_agents/
├── orchestrator_agent/
├── coder_agent/
├── tester_agent/
├── agent_messaging/
└── storage/

It understands:

• Agent discovery via agent.json files
• Message routing through orchestrator
• Transport layers (WebSocket, RAG board)
• Log locations and formats
• Common configuration patterns

Requirements

• Python 3.10+
• Rich library (for test script)
• Agent messaging system installed
• Access to agent processes and logs

Maintenance

Updating the Skill

To add new diagnostic capabilities:

1. Edit SKILL.md to add new workflow steps
2. Update common_issues.md with new issues and solutions
3. Extend test script if needed for new checks

Adding New Agents

When adding new agents to the system:

1. Update routing keyword mappings in examples
2. Add agent-specific checks to diagnostic workflow
3. Document common issues for that agent

Tips for Best Results

1. Be specific: "coder agent not responding" is better than "agent problem"
2. Provide context: Mention what you tried and what error you got
3. Share symptoms: "no response" vs "wrong response" lead to different diagnostics
4. Run test script: Provides baseline data for Claude to analyze

Troubleshooting the Skill Itself

If Claude doesn't load the skill when expected:

1. Check the skill description contains relevant keywords
2. Make sure SKILL.md has valid YAML frontmatter
3. Try more explicit requests: "Use the agent-debug skill to check status"
4. Verify file is at: .claude/skills/agent-debug/SKILL.md

Contributing

To improve this skill:

1. Add new common issues to common_issues.md
2. Enhance test script with more checks
3. Add examples for new scenarios
4. Document new agent types or configurations

Related Skills

• May want to create additional skills for:
  • Performance monitoring
  • Agent development/creation
  • Message board analysis
  • Transport layer debugging

---

Version: 1.0 Author: Claude Code Collective Last Updated: 2024-12-05