一键导入
debug
Unified debugging workflow enforcing root cause protocol. Investigate → Understand → Reason → Fix. Use when debugging bugs, errors, or unexpected behavior.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Unified debugging workflow enforcing root cause protocol. Investigate → Understand → Reason → Fix. Use when debugging bugs, errors, or unexpected behavior.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | debug |
| description | Unified debugging workflow enforcing root cause protocol. Investigate → Understand → Reason → Fix. Use when debugging bugs, errors, or unexpected behavior. |
Complete debugging workflow that enforces root cause analysis: Investigate → Understand → Reason → Fix.
Debug combines multiple debugging approaches:
# Gemini CLI for AI analysis
pip install google-generativeai
export GEMINI_API_KEY=your_api_key
# gh CLI for GitHub issue search
brew install gh
gh auth login
# Phase 1: Investigate codebase
rg -l "error pattern" --type ts
rg "relevant_function" -A 10 -B 5
# Phase 2: Search GitHub for known issues
gh search issues "error message" --repo owner/repo --limit 10
gh issue view 123 --repo owner/repo
# Phase 3: Reason about root cause
gemini -m pro -o text -e "" "Given this error and code context, what is the root cause?
Error: [error message]
Code:
\$(cat src/file.ts)
Investigation findings:
- [finding 1]
- [finding 2]
Apply first principles, systematic, and critical thinking frameworks."
# Phase 4: Generate fix
gemini -m pro -o text -e "" "Based on root cause: [cause]
Provide:
1. Specific code fix
2. Before/after code
3. How to verify
4. How to prevent"
# Skip deep reasoning, focus on investigation + fix
rg "error pattern" --type ts -A 5 -B 5
gemini -m pro -o text -e "" "Debug this quickly:
Problem: [description]
Error: [message]
Code:
\$(cat src/problematic-file.ts)
Give me the root cause and fix in 3 sentences."
# When you already know which files are involved
gemini -m pro -o text -e "" "Debug using these files:
Problem: [description]
File 1:
\$(cat src/file1.ts)
File 2:
\$(cat src/file2.ts)
Provide:
1. Root cause
2. Specific fix with line numbers
3. Verification steps"
# Parallel investigation of multiple sources
# Terminal 1: Codebase search
rg "error" --type ts --stats > /tmp/codebase.txt &
# Terminal 2: GitHub issues
gh search issues "error" --limit 20 > /tmp/github.txt &
# Terminal 3: Git history
git log --oneline --all -S "problematic_function" > /tmp/history.txt &
wait
# Synthesize findings
gemini -m pro -o text -e "" "Synthesize these diagnostic findings:
Codebase:
\$(cat /tmp/codebase.txt)
GitHub Issues:
\$(cat /tmp/github.txt)
Git History:
\$(cat /tmp/history.txt)
Provide unified diagnosis with:
1. Most likely root cause
2. Confidence level
3. Key evidence
4. Differential diagnosis
5. Recommended action"
Always follow this hierarchy:
| Symptom | BANNED Fix | REQUIRED Fix |
|---|---|---|
| Null error | if (x) { x.y } | Find why x is null |
| Timeout | Increase timeout | Find what's slow |
| Flaky test | Skip test | Find race condition |
| Type error | as any | Fix type hierarchy |
# Find error patterns
rg "throw.*Error" --type ts -A 3
# Find function definitions
rg "function functionName|const functionName" --type ts
# Find usages
rg "functionName\(" --type ts
# Find recent changes
git log --oneline -20 --all -- src/problematic/
git diff HEAD~5 -- src/problematic/
# Search issues
gh search issues "error message" --repo owner/repo --state all
# View issue details
gh issue view 123 --repo owner/repo --comments
# Search across multiple repos
for repo in repo1 repo2 repo3; do
gh search issues "error" --repo owner/$repo --limit 5
done
# Find logs first
find . -name "*.log" -type f
# Tail recent logs
tail -100 logs/app.log | grep -i error
# Search logs for patterns
grep -n "ERROR\|WARN" logs/*.log | tail -50
gemini -m pro -o text -e "" "Apply first principles to this bug:
Problem: [description]
Questions:
1. What is this code supposed to do?
2. What is it actually doing?
3. What assumptions are being made?
4. Which assumption is wrong?"
gemini -m pro -o text -e "" "Systematically analyze:
Problem: [description]
Walk through:
1. Input → What data enters?
2. Processing → What transformations?
3. State → What state changes?
4. Output → What comes out?
5. Where does expected diverge from actual?"
gemini -m pro -o text -e "" "Challenge assumptions:
Problem: [description]
For each potential cause:
- What evidence supports it?
- What evidence contradicts it?
- What would we expect to see if true?
- Can we rule it out?"
A debug session should produce:
## ROOT CAUSE
Single sentence identifying the actual cause.
## CONFIDENCE
high | medium | low
## FIX
Specific code changes:
- File: src/file.ts
- Line: 42
- Before: `oldCode()`
- After: `newCode()`
## VERIFICATION
How to verify the fix works:
1. Run test: `npm test -- specific.test.ts`
2. Manual check: [steps]
## PREVENTION
How to prevent this in the future:
- Add validation at boundary
- Add regression test
Always search before starting any work across all coding agent session histories (Claude Code, Codex, Cursor, Gemini CLI, Aider, ChatGPT) to find whatever we've discussed before.
Spawn 5 Opus subagents with randomly-generated distinct personas to debate a problem from multiple angles. Use when exploring UX decisions, architecture choices, or any decision that benefits from diverse perspectives arguing creatively.
Configure Karabiner-Elements keyboard remapping using Goku EDN syntax. Use when creating keybindings, layers, simlayers, app-specific shortcuts, or modifying karabiner.edn.
Bundle code context for AI. ALWAYS use --limit 49k unless user explicitly requests otherwise. Use for creating shareable code bundles and preparing context for LLMs.
Build Raycast extensions with React and TypeScript. Use when the user asks to create a Raycast extension, command, or tool.
Create new Agent Skills for Claude Code. Use when user wants to create a skill, add a new capability, document a CLI workflow, or asks how skills work.