# Install via npm
npm install -g @openai/codex

# Or via Homebrew (macOS)
brew install --cask codex

# Authenticate (recommended: ChatGPT account)
codex
# Follow authentication prompts

Code review request → What scope?
    ├─ Git changes (staged/unstaged) → Use: Git Diff Review
    │
    ├─ Pull Request → Use: PR Review Workflow
    │
    ├─ Specific files → Use: File Review
    │
    ├─ Entire directory/project → Use: Directory Review
    │
    └─ Special focus needed?
        ├─ Security concerns → Use: Security Audit
        ├─ Performance issues → Use: Performance Review
        └─ Architecture/Design → Use: Architecture Review

# CORRECT: Full automation mode - grants all permissions automatically
codex --full-auto exec "Review the staged git changes..."

# WRONG: May hang waiting for approval in automated contexts
codex exec "Review the staged git changes..."

# Auto-approve specific operation types
codex --auto-approve-read --auto-approve-execute exec "..."

codex --full-auto exec "Review the staged git changes. Analyze code quality, identify bugs, suggest improvements, and check for security issues. Provide a structured review with severity levels."

codex --full-auto exec "Review all staged changes (git diff --cached). For each file:
1. Summarize what changed
2. Identify potential bugs or logic errors
3. Check for security vulnerabilities
4. Suggest code quality improvements
5. Rate severity: critical/high/medium/low

Format as a structured review report."

codex --full-auto exec "Review all uncommitted changes (git diff HEAD). Provide:
- Summary of changes per file
- Bug identification with line numbers
- Security concerns
- Code style issues
- Suggested fixes with code examples"

codex --full-auto exec "Review changes between main and current branch (git diff main...HEAD). Focus on:
1. Breaking changes
2. API compatibility
3. Test coverage gaps
4. Documentation needs"

# First, fetch PR diff
gh pr diff <PR_NUMBER> > /tmp/pr_diff.txt

# Then review with codex (--full-auto for headless operation)
codex --full-auto exec "Review the code changes in /tmp/pr_diff.txt as a thorough PR reviewer. Provide:

## Summary
Brief description of what this PR accomplishes

## Code Review
For each file changed:
- Purpose of changes
- Potential issues (bugs, edge cases)
- Security considerations
- Performance implications

## Recommendations
- Required changes (blocking)
- Suggested improvements (non-blocking)
- Questions for the author

## Verdict
APPROVE / REQUEST_CHANGES / NEEDS_DISCUSSION"

codex --full-auto exec "Perform a comprehensive code review of src/utils/auth.ts. Analyze:
1. Code correctness and logic
2. Error handling completeness
3. Security vulnerabilities (OWASP Top 10)
4. Performance bottlenecks
5. Code maintainability
6. Test coverage recommendations"

codex --full-auto exec "Review these files as a cohesive unit: src/api/handler.ts, src/api/middleware.ts, src/api/routes.ts. Focus on:
- Consistency across files
- Proper separation of concerns
- Error propagation
- Request validation"

codex --full-auto exec "Perform a code review of the src/services/ directory. For each file:
- Identify the file's purpose
- List any bugs or issues
- Note security concerns
- Suggest improvements

Provide a summary with prioritized action items."

codex --full-auto exec "Perform a security audit of the codebase. Check for:

**Critical:**
- SQL injection vulnerabilities
- Command injection risks
- Authentication/authorization flaws
- Sensitive data exposure
- Insecure deserialization

**High:**
- XSS vulnerabilities
- CSRF issues
- Insecure dependencies
- Hardcoded secrets/credentials
- Improper input validation

**Medium:**
- Missing rate limiting
- Verbose error messages
- Insecure configurations
- Missing security headers

Report findings with:
- Severity level
- File and line number
- Description of vulnerability
- Remediation steps
- Code fix examples"

codex --full-auto exec "Analyze the codebase for performance issues:

1. **Algorithm Complexity**
   - O(n^2) or worse operations
   - Unnecessary nested loops
   - Inefficient data structures

2. **Resource Usage**
   - Memory leaks
   - Unclosed resources
   - Large object allocations

3. **I/O Operations**
   - N+1 query patterns
   - Synchronous blocking calls
   - Missing caching opportunities

4. **Concurrency**
   - Race conditions
   - Deadlock potential
   - Thread safety issues

Provide specific file locations and optimization suggestions."

codex --full-auto exec "Review the codebase architecture:

1. **Design Patterns**
   - Identify patterns in use
   - Suggest missing patterns
   - Flag anti-patterns

2. **SOLID Principles**
   - Single Responsibility violations
   - Open/Closed principle adherence
   - Dependency Inversion issues

3. **Code Organization**
   - Module boundaries
   - Circular dependencies
   - Coupling analysis

4. **Maintainability**
   - Code duplication
   - Complex functions (cyclomatic complexity)
   - Missing abstractions

Provide architectural recommendations with examples."

codex --full-auto exec --model gpt-5.1-codex "Perform thorough code review of src/..."

model_reasoning_effort = "high"

codex --full-auto exec -o review_report.md "Review src/api/..."

codex --full-auto exec --json "Review staged changes. Return JSON with structure:
{
  \"summary\": \"...\",
  \"files_reviewed\": [...],
  \"issues\": [{\"severity\": \"...\", \"file\": \"...\", \"line\": N, \"message\": \"...\"}],
  \"recommendations\": [...]
}" 2>/dev/null | jq '.item.content'

#!/bin/bash
# ci-review.sh

# Get changed files
CHANGED_FILES=$(git diff --name-only origin/main...HEAD)

if [ -z "$CHANGED_FILES" ]; then
    echo "No files changed"
    exit 0
fi

# Run codex review (--full-auto required for CI/headless operation)
codex --full-auto exec --skip-git-repo-check -o review.md "Review these changed files: $CHANGED_FILES

Provide a structured review. If any critical or high severity issues are found, clearly indicate BLOCKING_ISSUES=true at the end."

# Check for blocking issues
if grep -q "BLOCKING_ISSUES=true" review.md; then
    echo "Critical issues found in code review"
    cat review.md
    exit 1
fi

echo "Code review passed"
exit 0

Review the following code changes. For each issue found:

1. **Location**: File path and line number
2. **Severity**: Critical / High / Medium / Low / Info
3. **Category**: Bug / Security / Performance / Style / Documentation
4. **Description**: Clear explanation of the issue
5. **Suggestion**: How to fix it with code example

Organize by severity, starting with Critical issues.

As a senior engineer, review this PR for merge readiness:

## Checklist
- [ ] Code correctness verified
- [ ] No security vulnerabilities
- [ ] Performance acceptable
- [ ] Error handling complete
- [ ] Tests adequate
- [ ] Documentation updated

## Issues Found
[List any blocking or non-blocking issues]

## Verdict
[APPROVE / REQUEST_CHANGES with specific required changes]

# Use WebSearch tool to verify current information
# Example queries:
# - "ArgoCD latest version 2025"
# - "React 19 release date"
# - "Node.js LTS current version"
# - "[library name] latest stable release"

**Dependency Check** (verified via web search on YYYY-MM-DD):
- Package X: Using v1.2.3, latest stable is v2.0.1 ✓
- Package Y: Using v3.0.0, this IS the latest version ✓
- Package Z: Using v0.9.0, latest is v1.0.0 (breaking changes - review release notes)