An AI-powered GitHub Action that automatically detects spam in issues and comments using LLMs. Detected issues are tagged as spam and detected comments are (optionally) minimized.

Each new issue and comment is passed through a series of prompts: by default checking for link spam, regular spam, and AI-generated content, but you can write your own custom prompt if you've got more specific requirements.

Add this action to your repository's workflow file (e.g., .github/workflows/moderator.yml):

name: AI Moderator
on:
  issues:
    types: [opened]
  issue_comment:
    types: [created]
  pull_request_review_comment:
    types: [created]

jobs:
  spam-detection:
    runs-on: ubuntu-latest
    permissions:
      issues: write
      pull-requests: write
      models: read
      contents: read
    steps:
      - uses: actions/checkout@v4
      - uses: github/ai-moderator@v1
        with:
          token: ${{ secrets.GITHUB_TOKEN }}
          spam-label: 'spam'
          ai-label: 'ai-generated'
          minimize-detected-comments: true
          # Built-in prompt configuration (all enabled by default)
          enable-spam-detection: true
          enable-link-spam-detection: true
          enable-ai-detection: true
          # custom-prompt-path: '.github/prompts/my-custom.prompt.yml'  # Optional

The action does not require any external API keys for inference - it uses the built-in GitHub token with models: read permission to access GitHub Models.

Every GitHub user has GitHub Models inference for free, but if you're running into rate limiting issues you can choose to opt in to paid usage.

The action uses built-in YAML prompts located in the prompts/ directory. Each prompt can be individually enabled or disabled using the configuration options:

• spam-detection.prompt.yml: Detects promotional content, scams, and irrelevant posts (controlled by enable-spam-detection)
• ai-detection.prompt.yml: Identifies AI-generated content patterns (controlled by enable-ai-detection)
• link-spam-detection.prompt.yml: Focuses on suspicious links and URLs (controlled by enable-link-spam-detection)

All prompts are enabled by default. You can selectively disable them based on your repository's moderation needs:

- uses: github/ai-moderator
  with:
    token: ${{ secrets.GITHUB_TOKEN }}
    enable-spam-detection: true # Enable general spam detection
    enable-link-spam-detection: false # Disable link spam detection
    enable-ai-detection: true # Enable AI content detection

You can iterate on or tweak these prompts via the Models tab on this repository. If you want to push an update to this prompt, please also include updated test data so we can see the effect of the prompt update.

You can also provide your own custom prompt file in your repository using the custom-prompt-path input:

- uses: github/ai-moderator
  with:
    token: ${{ secrets.GITHUB_TOKEN }}
    custom-prompt-path: '.github/prompts/my-custom-spam-detection.prompt.yml'

Custom prompt files should follow the same YAML format as the built-in prompts. An example custom prompt file is included at .github/prompts/example-custom.prompt.yml that demonstrates the proper format and shows how to create repository-specific spam detection rules.

# Install dependencies
npm install

# Run tests
npm test

# Build the action
npm run package

# Run linting
npm run lint

The action includes comprehensive tests for all modules:

# Run all tests with coverage
npm run test

# Run tests in watch mode
npm run test:watch

This project is licensed under the terms of the MIT open source license. Please refer to LICENSE for the full terms.