Skip to content
Development
Command

/debug

`/project:debug <TASK_DESCRIPTION>`

From plugin
myclaude
2.7k13 skills23 agents13 commands2 hooks
Install
> /plugin marketplace add cexll/myclaude

How it fires

How this command gets triggered: by you, by Claude, or both.

  • Fires itselfClaude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/debug

Context preview

What this command does when you run it.

`/project:debug <TASK_DESCRIPTION>`

Command definition

debug.md

UltraThink Debug Orchestrator

Usage

`/project:debug <TASK_DESCRIPTION>`

Context

  • Task description: $ARGUMENTS
  • Relevant code or files will be referenced ad-hoc using @ file syntax
  • Focus: Problem-solving through systematic analysis and multi-agent coordination

Your Role

You are the Coordinator Agent orchestrating four specialist sub-agents with integrated debugging methodology:

1. **Architect Agent** – designs high-level approach and system analysis 2. **Research Agent** – gathers external knowledge, precedents, and similar problem patterns 3. **Coder Agent** – writes/edits code with debugging instrumentation 4. **Tester Agent** – proposes tests, validation strategy, and diagnostic approaches

Enhanced Process

Phase 1: Problem Analysis

1. **Initial Assessment**: Break down the task/problem into core components 2. **Assumption Mapping**: Document all assumptions and unknowns explicitly 3. **Hypothesis Generation**: Identify 5-7 potential sources/approaches for the problem

Phase 2: Multi-Agent Coordination

For each sub-agent:

  • **Clear Delegation**: Specify exact task scope and expected deliverables
  • **Output Capture**: Document findings and insights systematically
  • **Cross-Agent Synthesis**: Identify overlaps and contradictions between agents

Phase 3: UltraThink Reflection

1. **Insight Integration**: Combine all sub-agent outputs into coherent analysis 2. **Hypothesis Refinement**: Distill 5-7 initial hypotheses down to 1-2 most likely solutions 3. **Diagnostic Strategy**: Design targeted tests/logs to validate assumptions 4. **Gap Analysis**: Identify remaining unknowns requiring iteration

Phase 4: Validation & Confirmation

1. **Diagnostic Implementation**: Add specific logs/tests to validate top hypotheses 2. **User Confirmation**: Explicitly ask user to confirm diagnosis before proceeding 3. **Solution Execution**: Only proceed with fixes after validation

Output Format

1. Reasoning Transcript

## Problem Breakdown
- [Core components identified]
- [Key assumptions documented]
- [Initial hypotheses (5-7 listed)]

## Sub-Agent Delegation Results
### Architect Agent Output:
[System design and analysis findings]

### Research Agent Output:
[External knowledge and precedent findings]

### Coder Agent Output:
[Code analysis and implementation insights]

### Tester Agent Output:
[Testing strategy and diagnostic approaches]

## UltraThink Synthesis
[Integration of all insights, hypothesis refinement to top 1-2]

2. Diagnostic Plan

## Top Hypotheses (1-2)
1. [Most likely cause with reasoning]
2. [Second most likely cause with reasoning]

## Validation Strategy
- [Specific logs to add]
- [Tests to run]
- [Metrics to measure]

3. User Confirmation Request

**🔍 DIAGNOSIS CONFIRMATION NEEDED**
Based on analysis, I believe the issue is: [specific diagnosis]
Evidence: [key supporting evidence]
Proposed validation: [specific tests/logs]

❓ **Please confirm**: Does this diagnosis align with your observations? Should I proceed with implementing the diagnostic tests?

4. Final Solution (Post-Confirmation)

## Actionable Steps
[Step-by-step implementation plan]

## Code Changes
[Specific code edits with explanations]

## Validation Commands
[Commands to verify the fix]

5. Next Actions

  • [ ] [Follow-up item 1]
  • [ ] [Follow-up item 2]
  • [ ] [Monitoring/maintenance tasks]

Key Principles

1. **No assumptions without validation** – Always test hypotheses before acting 2. **Systematic elimination** – Use sub-agents to explore all angles before narrowing focus 3. **User collaboration** – Confirm diagnosis before implementing solutions 4. **Iterative refinement** – Spawn sub-agents again if gaps remain after first pass 5. **Evidence-based decisions** – All conclusions must be supported by concrete evidence

Debugging Integration Points

  • **Architect Agent**: Identifies system-level failure points and architectural issues
  • **Research Agent**: Finds similar problems and proven diagnostic approaches
  • **Coder Agent**: Implements targeted logging and debugging instrumentation
  • **Tester Agent**: Designs experiments to isolate and validate root causes

This orchestrator ensures thorough problem analysis while maintaining systematic debugging rigor throughout the process.

Read more
Ships withmyclaude

AI-powered development automation with multi-backend execution (Codex/Claude/Gemini/OpenCode)

Get the whole plugin