Claude Code Team Development Guide — AI Pair Programming with CLAUDE.md, Git Hooks, and Automated Reviews

#!/bin/bash
# .git/hooks/pre-commit (or .husky/pre-commit)
 
echo "🔍 Running Claude Code pre-commit check..."
 
# Get staged files
STAGED_FILES=$(git diff --cached --name-only --diff-filter=ACM | grep -E '\.(ts|tsx|js|jsx)$')
 
if [ -z "$STAGED_FILES" ]; then
  echo "✅ No files to check"
  exit 0
fi
 
# Run Claude Code quality check
claude -p "Review the following file changes.
Only output 'BLOCK' if there are critical issues
(security vulnerabilities, type errors, unhandled exceptions).
Output 'PASS' if there are no critical problems.
 
Files:
$STAGED_FILES
 
$(git diff --cached -- $STAGED_FILES)" 2>/dev/null | tail -1 | grep -q "BLOCK"
 
if [ $? -eq 0 ]; then
  echo "❌ Critical issues detected. Commit aborted."
  echo "   Run 'git diff --cached' to review your changes."
  exit 1
fi
 
echo "✅ Code quality check passed"
exit 0

#!/bin/bash
# .git/hooks/prepare-commit-msg
 
COMMIT_MSG_FILE=$1
COMMIT_SOURCE=$2
 
# Skip merge or squash commits
if [ "$COMMIT_SOURCE" = "merge" ] || [ "$COMMIT_SOURCE" = "squash" ]; then
  exit 0
fi
 
# Generate commit message from the staged diff
DIFF=$(git diff --cached --stat)
DETAILED_DIFF=$(git diff --cached | head -500)
 
SUGGESTED_MSG=$(claude -p "Generate a single-line commit message in
Conventional Commits format from the following git diff.
Keep it under 72 characters.
Use feat/fix/refactor/docs/test/chore as the type prefix.
 
Change summary:
$DIFF
 
Change details:
$DETAILED_DIFF" 2>/dev/null | tail -1)
 
if [ -n "$SUGGESTED_MSG" ]; then
  echo "$SUGGESTED_MSG" > "$COMMIT_MSG_FILE"
fi

# Use a project-level .githooks/ directory
git config core.hooksPath .githooks
 
# Add a postinstall script to package.json
# "postinstall": "git config core.hooksPath .githooks"

# .github/workflows/ai-code-review.yml
name: AI Code Review
on:
  pull_request:
    types: [opened, synchronize]
 
jobs:
  review:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      pull-requests: write
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
 
      - name: Get changed files
        id: changes
        run: |
          FILES=$(gh pr diff ${{ github.event.pull_request.number }} \
            --name-only | grep -E '\.(ts|tsx|js|jsx)$' | head -20)
          echo "files=$FILES" >> $GITHUB_OUTPUT
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
      - name: AI Review
        if: steps.changes.outputs.files != ''
        run: |
          # Use Claude API for the review (example implementation)
          DIFF=$(gh pr diff ${{ github.event.pull_request.number }})
          # Generate review comments and post to the PR
          echo "Review complete"
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}

## Code Review Standards (add to CLAUDE.md)
 
### Must-check (BLOCK verdict)
- Potential SQL injection
- XSS vulnerabilities (improper use of dangerouslySetInnerHTML)
- Missing authentication/authorization checks
- Hardcoded environment variables or secrets
- Potential infinite loops or recursion
 
### Recommended checks (WARNING verdict)
- Unnecessary re-renders
- Potential N+1 query patterns
- Missing error handling
- Magic numbers
- High cyclomatic complexity (> 10)

# Set explicit context at the start of a task
claude "Current task: Refactoring user authentication
Assignee: Frontend (backend API being worked on by @tanaka in parallel)
Goal: Migrate from JWT to Session Cookies
Constraint: Existing API response format must not change
Reference PR: #234 (backend changes)"

## Branch-specific Rules (add to CLAUDE.md)
 
### feature/* branches
- Focus on new feature implementation
- Encourage writing tests alongside code
- Enforce TypeScript strict mode
 
### hotfix/* branches
- Keep changes minimal
- Adding related tests is mandatory
- Always verify performance impact
 
### release/* branches
- No new features allowed
- Only bug fixes and documentation updates