Files
gh-taylorhuston-ai-toolkit-…/commands/troubleshoot.md
2025-11-30 09:00:21 +08:00

111 lines
3.4 KiB
Markdown

---
tags: ["debugging", "investigation", "problem-solving"]
description: "5-step debugging loop for unexpected issues during implementation or standalone bug fixes"
argument-hint: "[BUG-### | TASK-### | \"description\"]"
allowed-tools: ["Read", "Write", "Edit", "Bash", "Grep", "Glob", "WebSearch", "WebFetch", "Task"]
model: claude-sonnet-4-5
references_guidelines:
- docs/development/workflows/bug-workflow.md # 5-step debugging loop, reproduction-first approach
- docs/development/workflows/worklog-format.md # WORKLOG entry formats
---
# /troubleshoot Command
**WHAT**: Systematic debugging using the 5-step loop from bug-workflow.md.
**WHY**: Avoid shotgun debugging. Research first, one hypothesis at a time, validate before claiming fixed.
**HOW**: Follow the 5-step debugging loop in bug-workflow.md.
## Usage
```bash
/troubleshoot "tests failing after auth changes" # During TASK - describe the issue
/troubleshoot TASK-001 # Debug issue in specific task context
/troubleshoot BUG-007 # Equivalent to /implement BUG-007
```
## When to Use
| Situation | Command |
|-----------|---------|
| Unexpected issue during TASK implementation | `/troubleshoot "description"` or `/troubleshoot TASK-###` |
| Known bug with BUG.md created | `/troubleshoot BUG-###` or `/implement BUG-###` |
| Tests failing, unclear why | `/troubleshoot` |
## Execution
### Mode 1: During TASK Implementation
When something unexpected breaks while implementing a TASK:
1. **Read context**: Current TASK.md, PLAN.md, WORKLOG.md
2. **Execute 5-step loop** from bug-workflow.md
3. **Document** in current WORKLOG.md
4. **Continue** with TASK implementation
### Mode 2: Standalone Bug (BUG-###)
Equivalent to `/implement BUG-###`:
1. **Read context**: BUG.md, PLAN.md, WORKLOG.md
2. **Execute 5-step loop** from bug-workflow.md
3. **Document** in WORKLOG.md with root cause
4. **Complete** with `/complete BUG-###`
## The 5-Step Loop
**See bug-workflow.md for complete details.**
```
Research → Hypothesize → Implement → Test → Document
↑ │
└─────────── (if not fixed) ──────────────┘
```
### Quick Reference
1. **Research** - Context7 → docs → ask user (don't guess)
2. **Hypothesize** - ONE theory with debug plan
3. **Implement** - Fix + liberal debug logging
4. **Test** - Run tests, verify logs, confirm fix
5. **Document** - WORKLOG entry with hypothesis/findings/result
### Key Rules
- Research BEFORE guessing
- ONE hypothesis at a time
- Liberal debug logging (`console.log('[Component] State:', data)`)
- NEVER claim "fixed" without tests passing
- Rollback on failure before next attempt
## WORKLOG Entry
```markdown
## 2025-11-26 - Troubleshooting Loop 1
**Hypothesis:** Query fires before auth completes
**Debug findings:**
- isLoading was true when query executed
- Auth state not propagating to component
**Implementation:** Added auth state check
**Result:** ✓ Fixed - 47/47 tests passing
```
## Error Handling
| Situation | Action |
|-----------|--------|
| Hypothesis fails | Rollback, document, return to Step 1 |
| 3+ failed loops | Pause, ask user for context |
| Tests pass but issue persists | More logging, manual verification |
## Integration
**During TASK:** Documents in current WORKLOG, returns to `/implement`
**For BUG:** Same as `/implement BUG-###`, follows bug-workflow.md