investigate
使用搜索和AI分析调试和调查代码问题。在遇到bug障碍、追踪执行流程或理解复杂代码时使用。
📖 中文导读:使用搜索和AI分析调试和调查代码问题。在遇到bug障碍、追踪执行流程或理解复杂代码时使用。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: investigate description: Debug and investigate code issues using search and AI analysis. Use when stuck on bugs, tracing execution flow, or understanding complex code.
Investigation Toolkit
Debug issues through systematic search and AI-powered analysis.
Prerequisites
# ripgrep for fast search
brew install ripgrep
# Gemini for analysis
pip install google-generativeai
export GEMINI_API_KEY=your_api_key
Search Commands
ripgrep Basics
# Search for pattern
rg "pattern" src/
# Case insensitive
rg -i "error" src/
# Whole word
rg -w "user" src/
# File types
rg -t ts "function" src/
rg -t py "def " src/
# Exclude patterns
rg "TODO" --glob "!node_modules"
# Show context
rg -C 3 "error" src/ # 3 lines before and after
rg -B 5 "crash" src/ # 5 lines before
rg -A 5 "crash" src/ # 5 lines after
# Just filenames
rg -l "pattern" src/
# Count matches
rg -c "pattern" src/
Finding Definitions
# Function definitions (TypeScript)
rg "function\s+functionName" src/
rg "(const|let|var)\s+functionName\s*=" src/
rg "export\s+(async\s+)?function\s+\w+" src/
# Class definitions
rg "class\s+ClassName" src/
# Interface/Type definitions
rg "(interface|type)\s+TypeName" src/
Tracing Usage
# Where is this function called?
rg "functionName\(" src/
# Where is this imported?
rg "import.*functionName" src/
# Where is this exported?
rg "export.*functionName" src/
Investigation Patterns
Bug Investigation
# 1. Search for error message
rg "exact error message" .
# 2. Find where error is thrown
rg "throw.*Error" src/ -C 3
# 3. Trace the function
rg "functionThatFails" src/ -C 5
# 4. Check recent changes
git log --oneline -20 --all -- src/problematic-file.ts
git diff HEAD~5 -- src/problematic-file.ts
Trace Execution Flow
#!/bin/bash
ENTRY_POINT=$1
echo "=== Entry Point ==="
rg -A 10 "export.*$ENTRY_POINT" src/
echo "=== Called Functions ==="
rg -o "\w+\(" src/$ENTRY_POINT*.ts | sort -u
echo "=== Dependencies ==="
rg "^import" src/$ENTRY_POINT*.ts
AI-Assisted Debugging
# Analyze error with context
ERROR="Your error message here"
CODE=$(cat problematic-file.ts)
gemini -m pro -o text -e "" "Debug this error:
ERROR: $ERROR
CODE:
$CODE
Provide:
1. Most likely cause
2. How to verify
3. How to fix
4. How to prevent in future"
Hypothesis Testing
# Generate hypotheses
gemini -m pro -o text -e "" "Given this bug symptom:
SYMPTOM: [describe what's happening]
CONTEXT: [relevant code/system info]
Generate 5 hypotheses ranked by likelihood, with a test for each."
# Then test each hypothesis
rg "hypothesis-related-pattern" src/
Common Investigations
Find All Error Handling
rg "catch|\.catch|try\s*{" src/ -t ts
rg "throw\s+new" src/ -t ts
Find API Endpoints
rg "(get|post|put|delete|patch)\s*\(" src/ -i
rg "router\.(get|post|put|delete)" src/
rg "@(Get|Post|Put|Delete)" src/
Find Database Queries
rg "(SELECT|INSERT|UPDATE|DELETE)" src/ -i
rg "\.query\(|\.execute\(" src/
rg "prisma\.\w+\.(find|create|update|delete)" src/
Find Configuration
rg "process\.env\." src/
rg "(config|settings)\[" src/
rg "getenv|os\.environ" src/ -t py
Find Security Issues
# SQL injection potential
rg "query.*\+.*\"|'.*\+" src/
# Hardcoded secrets
rg "(password|secret|key|token)\s*=\s*['\"]" src/ -i
# Unsafe eval
rg "eval\(" src/
Deep Investigation Script
#!/bin/bash
# investigate.sh - Comprehensive code investigation
TERM=$1
echo "=== Investigating: $TERM ==="
echo ""
echo "### Definitions ###"
rg "^(export\s+)?(function|const|class|interface|type)\s+$TERM" src/
echo ""
echo "### Usage ###"
rg "$TERM" src/ --stats | head -50
echo ""
echo "### Recent Changes ###"
git log --oneline -10 -S "$TERM"
echo ""
echo "### Blame ###"
for f in $(rg -l "$TERM" src/); do
echo "--- $f ---"
git blame -L "/$TERM/,+5" "$f" 2>/dev/null | head -10
done
Best Practices
- Start with the error - Search for exact message first
- Expand context - Use
-C,-B,-Afor surrounding code - Check history -
git log -Sfinds when code was introduced - Use AI for complex - When pattern matching isn't enough
- Document findings - Note what you discover
- Test hypotheses - Verify before assuming