redpen

Code review checklist for AI-generated code. Step through prompts, paste into your AI, fix issues, mark done.

Overview

redpen is a CLI tool that guides you through a set of code review prompts. Each prompt tells an AI what to look for in your codebase. You copy the prompt, paste it into ChatGPT/Claude/Cursor, fix what it finds, then mark it done. Progress saves per git branch.

Features

• Interactive TUI: Arrow keys to navigate, single-key actions
• First-Run Setup: Auto-detects your stack, guided init wizard
• Fuzzy Search: Press / to quickly find prompts
• Progress Tracking: Saves per git branch, pick up where you left off
• Stack-Aware: Loads prompts for your specific stack (Next.js, Supabase, Flutter, etc.)
• Custom Prompts: Add your own prompts in .redpen/ folder
• CI Integration: Fail builds if prompts incomplete

Installation

npm install -g redpen

Usage

# Start the TUI (runs init wizard on first use)
redpen

# Or explicitly run setup
redpen init

Keyboard Shortcuts

| Key | Action | |-----|--------| | r / Enter | Run: copy + mark done + advance | | c | Copy only (no mark done) | | s | Skip prompt + advance | | n | Jump to next incomplete | | d | Toggle done / undo | | / | Search prompts (fuzzy) | | ? | Help overlay | | ← → h l | Previous / next prompt | | ↑ ↓ k j | Previous / next prompt | | Home End | First / last prompt | | q / Esc | Quit |

Search Mode (/)

• Type to fuzzy filter prompts
• ↑ ↓ to navigate results
• Enter to select
• Esc to cancel

How It Works

1. Run redpen - TUI opens at first incomplete prompt
2. Press r - copies prompt, marks done, advances to next
3. Paste into your AI - it reviews your code
4. Fix what it finds
5. Repeat until complete

One key does it all - no more r then d then arrow key.

Commands

| Command | Description | |---------|-------------| | redpen | Interactive TUI (recommended) | | redpen init | First-time setup | | redpen next | Show next prompt | | redpen copy [n] | Copy prompt n | | redpen done [n] | Mark prompt n complete | | redpen skip [n] | Skip prompt n | | redpen undo | Undo last done | | redpen status | Show progress | | redpen list | List all prompts | | redpen order [tag] | Show run order | | redpen reset | Clear progress | | redpen check [cats] | CI mode - exit 1 if incomplete | | redpen report | Markdown summary |

Prompts

Core (always loaded)

| Category | What it checks | |----------|----------------| | core/security/* | Auth, data safety, access control | | core/quality/* | Test coverage, error handling | | core/architecture/* | Logging, monitoring, observability | | core/process/* | Documentation, code standards |

Stack-Specific (based on config)

| Stack | Prompts | |-------|---------| | Next.js | web/frontend/nextjs/* | | React | web/frontend/react/* | | Vue | web/frontend/vue/* | | Supabase | web/backend/supabase/* | | Firebase | web/backend/firebase/* | | Prisma | web/backend/prisma/* | | Flutter | mobile/flutter/* | | React Native | mobile/react-native/* | | Native iOS/Android | mobile/native/* |

Custom Prompts

Add .txt files to .redpen/ in your project root:

your-project/
  .redpen/
    my-custom-check.txt
    team/coding-standards.txt

They get added to the run order automatically.

CI Integration

# GitHub Actions
- run: npx redpen check security,quality

Exits with code 1 if security or quality prompts are incomplete.

Configuration

Run redpen init or just run redpen - the TUI will guide you through setup on first use.

Config is saved to .redpenrc:

{
  "platform": "web",
  "frontend": "nextjs",
  "backend": "supabase"
}

For mobile:

{
  "platform": "mobile",
  "framework": "flutter"
}

Auto-Detection

redpen automatically detects your stack from:

| File | Detection | |------|-----------| | pubspec.yaml | Flutter | | react-native in package.json | React Native | | next in package.json | Next.js | | @supabase/supabase-js | Supabase | | firebase | Firebase | | @prisma/client | Prisma |

Detected values are pre-selected in the init wizard.

Shell Completion

# Add to your shell profile
eval "$(redpen completion)"

Why This Exists

AI writes code that compiles and looks correct. But it misses things: security holes, missing error handling, edge cases, no tests, production gotchas.

These prompts catch what AI misses. Run them before you ship.

License

MIT