AgentSkillsCN

debug

启动系统化调试会话，采用结构化诊断协议

SKILL.md
--- frontmatter
name: debug
description: Start a systematic debugging session with structured diagnostic protocol
argument-hint: "<error or symptom description>"

Debugging Session

Start a systematic debugging session following a structured diagnostic protocol.

What This Skill Does

  1. •Gathers symptoms and error information
  2. •Forms hypotheses based on evidence
  3. •Tests systematically to isolate root cause
  4. •Documents findings for future reference

Initial Interaction

code
Let's debug this systematically. Tell me:

1. **What's happening?** (Symptom/error message)
2. **What should happen?** (Expected behavior)
3. **When did it start?** (Recent changes?)
4. **How often?** (Always / Sometimes / Once)

What's the error or symptom you're seeing?

Debugging Protocol

Phase 1: Gather Evidence

  • •Exact error message (copy/paste)
  • •Stack trace if available
  • •Steps to reproduce
  • •Environment details (dev/prod, versions)
  • •Recent changes (git log, deploys)

Phase 2: Form Hypotheses

Based on evidence, list likely causes:

  1. •Most likely: [hypothesis]
  2. •Possible: [hypothesis]
  3. •Less likely: [hypothesis]

Phase 3: Test Hypotheses

For each hypothesis:

  • •Define test that would confirm/reject
  • •Execute test
  • •Record result
  • •Move to next if rejected

Phase 4: Fix and Verify

  • •Implement fix
  • •Verify fix resolves issue
  • •Check for regressions
  • •Document root cause

Output Format

markdown
# Debugging Session: [Brief Description]

**Started**: [timestamp]
**Status**: Investigating / Root cause found / Resolved

---

## Symptoms

**Error**:

[Exact error message]

code

**Expected**: [What should happen]
**Actual**: [What happens instead]
**Reproducibility**: Always / Sometimes / Once
**Environment**: [Dev/Staging/Prod, versions]

---

## Recent Changes

| When | What Changed | Relevant? |
|------|--------------|-----------|
| [Date] | [Change] | Yes/No/Maybe |

---

## Hypotheses

### H1: [Most likely cause] - [TESTING/REJECTED/CONFIRMED]
**Why suspect**: [Reasoning]
**Test**: [How to verify]
**Result**: [What we found]

### H2: [Second possibility] - [TESTING/REJECTED/CONFIRMED]
**Why suspect**: [Reasoning]
**Test**: [How to verify]
**Result**: [What we found]

### H3: [Third possibility] - [TESTING/REJECTED/CONFIRMED]
**Why suspect**: [Reasoning]
**Test**: [How to verify]
**Result**: [What we found]

---

## Investigation Log

### [Timestamp] - [Action taken]
**Finding**: [What we learned]
**Next**: [What to try next]

### [Timestamp] - [Action taken]
**Finding**: [What we learned]
**Next**: [What to try next]

---

## Root Cause

**Identified**: [Yes/No/Partial]

[Description of root cause]

**Evidence**:
- [Evidence 1]
- [Evidence 2]

---

## Fix

**Approach**: [How we'll fix it]

**Changes**:
- [File:line] - [Change description]

**Verification**:
- [ ] Fix applied
- [ ] Original error no longer occurs
- [ ] No regressions introduced
- [ ] Tests pass

---

## Learnings

**Root Cause Pattern**: [What caused this]
**Prevention**: [How to prevent in future]
**Detection**: [How to catch earlier next time]

---

## Time Spent

| Phase | Duration |
|-------|----------|
| Gathering evidence | Xm |
| Forming hypotheses | Xm |
| Testing | Xm |
| Fixing | Xm |
| **Total** | **Xm** |

Common Debug Patterns

"It works on my machine"

  • •Check environment variables
  • •Check dependency versions
  • •Check file permissions
  • •Check network/firewall

"It worked yesterday"

  • •git log --since="yesterday"
  • •Check deploy history
  • •Check dependency updates
  • •Check external service changes

"It works sometimes"

  • •Race condition?
  • •Caching issue?
  • •External dependency flakiness?
  • •Resource limits?

"No error, just wrong"

  • •Check input data
  • •Check business logic
  • •Check edge cases
  • •Add logging to trace flow

Tools to Use

Quick Checks

bash
# Recent changes
git log --oneline -20

# What changed in a file
git diff HEAD~5 -- path/to/file

# Find where something is used
grep -r "functionName" --include="*.py"

# Check running processes
ps aux | grep [process]

# Check ports
lsof -i :8000

Add Debugging Output

  • •Console.log / print statements
  • •Structured logging
  • •Debugger breakpoints
  • •Network inspection

When to Escalate

  • •Debugging > 30 minutes without progress
  • •Issue involves external services you can't access
  • •Security implications discovered
  • •Need access/permissions you don't have

Post-Debug

After resolving, consider:

  • • Add test to prevent regression
  • • Document root cause in code comment
  • • Update monitoring/alerting
  • • Share learning with team
  • • /remember the pattern for future