majiayu000 Claude Skill: Agent System Manager
Skill Description
Manages and monitors the autonomous agent system for code implementation tasks. Provides progress checking, issue diagnosis, and automatic fixing capabilities.
Commands
1. Check Progress
/Users/mchaouachi/agent-system/check-agent-progress.sh
Shows complete status including:
- Current agent status (idle/awaiting_review/approved/implementing)
- Test progress (X/183 tests passing)
- Recent activity from logs
- Files recently modified
- Agent communication status
- Recommendations for next steps
2. Auto-Fix Issues
/Users/mchaouachi/agent-system/agent-autofix.sh /Users/mchaouachi/IdeaProjects/StockMonitor
Automatically diagnoses and fixes common issues:
- Creates proposals if missing
- Triggers reviewer if stuck
- Starts implementation if approved but not running
- Fixes communication issues between agents
- Only restarts as last resort
3. Quick Status
cat /Users/mchaouachi/IdeaProjects/StockMonitor/coordination/task_proposals.json | python3 -c "import json,sys; d=json.load(sys.stdin); print(f'Status: {d.get(\"status\")}, Proposals: {len(d.get(\"proposals\", []))}, Chosen: {d.get(\"chosen_approach\", \"none\")}')"
4. Force Implementation
tmux send-keys -t agent_system_spec:planner C-c Enter && sleep 2 && tmux send-keys -t agent_system_spec:planner "claude" Enter && sleep 3 && tmux send-keys -t agent_system_spec:planner "Read /Users/mchaouachi/IdeaProjects/StockMonitor/coordination/task_proposals.json. Implement the approved approach to fix 75 tests. Work autonomously." Enter
5. View Live Agents
tmux attach -t agent_system_spec
Then use:
- Ctrl+b 0 for planner
- Ctrl+b 1 for reviewer
- Ctrl+b 2 for monitor
6. Check Test Results
cd /Users/mchaouachi/IdeaProjects/StockMonitor && mvn test | grep "Tests run:"
7. Emergency Restart
tmux kill-session -t agent_system_spec && cd /Users/mchaouachi/agent-system && ./launch-agents-from-spec.sh /Users/mchaouachi/IdeaProjects/StockMonitor 999
Usage Examples
When to use each command:
Nothing happening?
/Users/mchaouachi/agent-system/check-agent-progress.sh
Then follow recommendations.
Agents stuck?
/Users/mchaouachi/agent-system/agent-autofix.sh
Automatically diagnoses and applies fixes.
Want to see what agents are doing?
tmux attach -t agent_system_spec
Need test status?
cd /Users/mchaouachi/IdeaProjects/StockMonitor && mvn test | grep "Tests run:"
File Locations
- Project:
/Users/mchaouachi/IdeaProjects/StockMonitor - Agent System:
/Users/mchaouachi/agent-system - Proposals:
/Users/mchaouachi/IdeaProjects/StockMonitor/coordination/task_proposals.json - Logs:
/Users/mchaouachi/IdeaProjects/StockMonitor/coordination/logs/ - Specs:
/Users/mchaouachi/IdeaProjects/StockMonitor/specs/
Status Meanings
- idle: No activity, needs to start
- awaiting_review: Proposals created, reviewer should evaluate
- approved: Ready for implementation
- implementing: Currently fixing tests
Troubleshooting Quick Guide
| Problem | Solution |
|---|---|
| No proposals | Run agent-autofix.sh |
| Stuck on "awaiting_review" | Run agent-autofix.sh |
| Approved but not implementing | Run agent-autofix.sh |
| Want fresh start | Kill session, delete task_proposals.json, restart |
| Need to see errors | Check planner window in tmux |
Integration with Claude
When called from Claude, I can:
- Run check-agent-progress.sh to get current status
- Analyze the output and recommend next steps
- Run agent-autofix.sh if issues detected
- Monitor progress and report back
- Provide specific commands to fix identified issues
This skill enables Claude to act as a supervisor for the autonomous agent system, ensuring smooth operation and fixing issues automatically.