🔥 Karen GitHub Action

Get brutally honest, AI-powered code reviews from Karen - no BS, just reality checks.

Works with Anthropic Claude OR OpenAI GPT-4 - Choose your preferred AI provider.

Features

• Auto-Update README Badge - Karen automatically keeps your score badge fresh
• Brutally Honest Reviews - AI-powered analysis that tells it like it is
• Karen Score (0-100) - Quantified assessment across 5 key dimensions
• Auto PR Comments - Karen roasts your PRs automatically
• Auto-Commit Results - Push scores, reviews, and badges back to your repo

Quick Start

1. Add markers to your README (optional, but recommended)

# My Project

<!-- karen-badge-start -->
<!-- karen-badge-end -->

Description...

2. Create .github/workflows/karen.yml

Default: Manual Trigger

name: Karen Code Review
on:
  workflow_dispatch:  # Manual trigger via Actions tab

jobs:
  karen-review:
    runs-on: ubuntu-latest
    permissions:
      contents: write
      pull-requests: write

    steps:
      - uses: actions/checkout@v4

      - name: Karen Review
        uses: khaliqgant/karen-action@v1
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          auto_update_readme: true  # Automatically update badge in README
          generate_badge: true
          post_comment: true

      - name: Commit results
        run: |
          git config user.name "github-actions[bot]"
          git config user.email "github-actions[bot]@users.noreply.github.com"
          git add .karen/ README.md
          git diff --staged --quiet || git commit -m "Update Karen review [skip ci]"
          git push

Optional: Auto-run on PRs

name: Karen Code Review
on:
  workflow_dispatch:  # Manual trigger
  pull_request:       # Auto-run on PRs
    branches: [main]

jobs:
  karen-review:
    runs-on: ubuntu-latest
    permissions:
      contents: write
      pull-requests: write

    steps:
      - uses: actions/checkout@v4

      - name: Karen Review
        uses: khaliqgant/karen-action@v1
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          auto_update_readme: true
          generate_badge: true
          post_comment: true

      - name: Commit results
        run: |
          git config user.name "github-actions[bot]"
          git config user.email "github-actions[bot]@users.noreply.github.com"
          git add .karen/ README.md
          git diff --staged --quiet || git commit -m "Update Karen review [skip ci]"
          git push

3. Add API Key

Get an API key and add it to your repository secrets:

• Anthropic Claude (Recommended): console.anthropic.com → Add as ANTHROPIC_API_KEY
• OpenAI GPT-4: platform.openai.com → Add as OPENAI_API_KEY

4. Get Roasted

Push code and Karen will:

• Analyze your repository
• Generate a score (0-100)
• Auto-update your README badge (clickable link to review)
• Comment on PRs
• Save detailed reviews to .karen/review.md

---

Karen Score

Karen evaluates across 5 dimensions (0-20 points each):

Dimension	What It Measures
🎭 Bullshit Factor	Over-engineering vs. simplicity
⚙️ Actually Works	Does it do what it claims?
💎 Code Quality	Will the next dev curse you?
✅ Completion	TODO vs. actually done
🎯 Practical Value	Solves real problems vs. resume padding

Score Grades:

• 90-100 🏆 Surprisingly legit
• 70-89 ✅ Actually decent
• 50-69 😐 Meh, it works I guess
• 30-49 🚨 Needs intervention
• 0-29 💀 Delete this and start over

Configuration

Input	Default	Description
anthropic_api_key	-	Anthropic API key for Claude
openai_api_key	-	OpenAI API key for GPT-4
auto_update_readme	false	Auto-insert/update badge in README
generate_badge	true	Generate Karen score badge
post_comment	true	Post review as PR comment
min_score	0	Minimum score threshold (shows warning if below)

Optional: .karen/config.yml

strictness: 8  # 1-10, how brutal Karen should be
weights:
  bullshitFactor: 0.25
  actuallyWorks: 0.25
  codeQualityReality: 0.20
  completionHonesty: 0.15
  practicalValue: 0.15

Examples

Minimal Setup

- uses: khaliqgant/karen-action@v1
  with:
    anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}

With Auto-Update Badge

- uses: khaliqgant/karen-action@v1
  with:
    anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
    auto_update_readme: true  # Badge updates automatically!

Enforce Minimum Score

- uses: khaliqgant/karen-action@v1
  with:
    anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
    min_score: 70  # Shows warning if score < 70

Using OpenAI

- uses: khaliqgant/karen-action@v1
  with:
    openai_api_key: ${{ secrets.OPENAI_API_KEY }}

What Gets Created

.karen/
├── score.json              # Current score & breakdown
├── review.md               # Latest review
├── history/                # Historical reviews
│   └── 2025-10-18-12-30.md
└── badges/                 # Generated badges
    └── score-badge.svg

💡 Want Karen in Your IDE?

Get Karen interactively in Claude Code or Cursor with PRPM:

npm install -g prpm
prpm install @prpm/karen-repo-reviewer-skill # Claude Code
prpm install @prpm/karen-repo-reviewer # Cursor IDE

FAQ

Q: Will Karen roast my code? A: Yes. That's literally her job.

Q: How does auto-update badge work? A: Karen finds <!-- karen-badge-start --> markers in your README (or adds them if missing) and updates the badge automatically. The badge is clickable and links to .karen/review.md so readers can see Karen's full hot take.

Q: Will this fail my CI? A: Only if you set min_score and Karen's score is below it.

Q: How much does this cost? A: You need an Anthropic or OpenAI API key. Typical review costs ~$0.10-0.50 depending on repo size.

License

MIT

---

🔥 Get roasted. Get better. - Karen