CodeAudits Action

Overview

This action uses the Repomix library to generate a comprehensive text representation of your codebase and optionally submits it to Google Gemini with selected prompt. This allows you to:

Create a single document containing your entire codebase in a format suitable for LLM analysis
Analyze your code with predefined or custom prompts through Google Gemini

What's New in v2.0.0

🤖 AI-Powered Code Audits: Integrate with Google Gemini for intelligent code analysis and insights
📝 Default Analysis Prompts: 8 pre-built prompts covering architecture, SOLID principles, security, testing, and more
🎯 Custom Prompt Support: Create and use your own custom analysis prompts from .codeaudits/prompts directory
💻 Local Development: Run the tool locally outside of GitHub Actions for faster development cycles
⚡ Updated Dependencies: Upgraded to Repomix v1.2.0 for improved parsing performance and reliability
📚 Enhanced Examples: More comprehensive GitHub Actions configuration examples and use cases

Usage

Add the following to your GitHub Actions workflow file:

name: CodeAudits Analysis

on:
  push:
    branches: [ main ]
  pull_request:
    branches: [ main ]
  # Or use workflow_dispatch to trigger manually
  workflow_dispatch:

jobs:
  analyze:
    runs-on: ubuntu-latest
    name: Analyze code with CodeAudits
    
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
        with:
          fetch-depth: 0  # Fetch all history for thorough analysis
      
      - name: Parse and submit repository
        uses: codeaudits/codeaudits-action@v1
        id: audit
        with:
          style: markdown            # Format of the parsed output
          compress: true             # Enable intelligent code parsing to reduce tokens (default: false)
          llm-prompt: architecture-refactoring  # Optional: AI analysis prompt
          gemini-api-key: ${{ secrets.GEMINI_API_KEY }}  # Optional: for AI analysis
      
      - name: Upload parsed file as artifact
        uses: actions/upload-artifact@v4
        with:
          name: codebase-analysis
          path: parsed-repo.txt

Inputs

Input	Description	Default	Required
`style`	Parsed document style. Use 'markdown', 'xml' or 'plain' (it's Repomix config option)	`markdown`	No
`compress`	Run intelligent code parsing to reduce tokens (it's Repomix config option)	`false`	No
`working-directory`	The directory in which to run the action. Defaults to the repository root	`.`	No
`llm-prompt`	The name of the prompt file for the Gemini prompt (see Available Prompts)		No
`llm-custom-prompt`	The name of the custom prompt file from `/.codeaudits/prompts` directory in your repository		No
`gemini-api-key`	Gemini API key for AI-powered code analysis		No

Available Prompts

When using the llm-prompt input, you can choose from the following predefined prompt files:

architecture-refactoring - Analysis focused on architectural improvements and refactoring opportunities
dry-kiss-yagni - Review based on DRY, KISS, and YAGNI principles
essential-software-patterns - Analysis of software design patterns usage
functionalities-analysis - Comprehensive functionality and feature analysis
missing-tests - Identification of missing test coverage and testing opportunities
possible-bugs - Detection of potential bugs and code issues
simplification-hints - Suggestions for code simplification and optimization
solid - Review based on SOLID principles

Note: If you provide an invalid prompt name, the action will fail with a helpful error message listing all available options.

Custom Prompts

In addition to the predefined prompts, you can create and use custom prompts by:

Creating a /.codeaudits/prompts directory in your repository
Adding your custom prompt file (e.g., my-custom-analysis.md)
Using the llm-custom-prompt input with the filename

Example custom prompt file structure:

your-repository/
├── .codeaudits/
│   └── prompts/
│       ├── performance-analysis.md
│       ├── security-review.md
│       └── api-design-review.md
└── ... (your code)

Usage with custom prompt:

- name: Run Custom AI Analysis
  uses: codeaudits/codeaudits-action@v1
  with:
    style: markdown
    llm-custom-prompt: performance-analysis.md  # or just: performance-analysis
    gemini-api-key: ${{ secrets.GEMINI_API_KEY }}

Important notes:

You cannot use both llm-prompt and llm-custom-prompt at the same time
The .md extension is optional when specifying the filename
Custom prompt files should contain markdown-formatted prompts for the AI analysis

Examples

Basic Analysis

- name: Run CodeAudits Analysis
  uses: codeaudits/codeaudits-action@v1
  with:
    style: markdown

AI-Powered Code Analysis

- name: Run AI-Powered Code Analysis
  uses: codeaudits/codeaudits-action@v1
  with:
    style: markdown
    compress: true
    llm-prompt: architecture-refactoring
    gemini-api-key: ${{ secrets.GEMINI_API_KEY }}

AI-Powered Code Analysis with Custom Prompts

- name: Run Custom AI Analysis
  uses: codeaudits/codeaudits-action@v1
  with:
    style: markdown
    compress: true
    llm-custom-prompt: security-review
    gemini-api-key: ${{ secrets.GEMINI_API_KEY }}

Setting up Gemini API Key:

Get your API key from Google AI Studio
Add it as a repository secret named GEMINI_API_KEY
The AI analysis results will appear in the GitHub Actions job summary

Example Workflows

You can find complete workflow examples in the .github/workflows directory:

llm_audit.yml - Complete workflow with AI-powered audit using predefined prompts
llm_audit_custom_prompt.yml - AI audit using custom prompts from your repository
llm_audit_external_codebase.yml - Audit external repositories with AI analysis
llm_parse_only.yml - Parse repository without AI analysis (output only)
llm_parse_only_external_codebase.yml - Parse external repositories without AI analysis

These examples demonstrate different use cases and can be copied directly to your repository's .github/workflows directory.

Viewing Results

After the action completes:

The parsed repository will be available as a GitHub Actions artifact named parsed-repo.txt
Detailed metadata about the parsing process will be available in the job outputs

License

This project is licensed under the MIT License - see the LICENSE file for details.

frogermcs / codeaudits-parse