- Use the Goose AI agent to automate creation of a GitHub Action
- The Action should download a repo, analyze the diff in a PR, and add relevant comments
- Minimize direct coding by leveraging Goose's capabilities
- Evaluate Goose features and extensibility model
- Understand how to customize with our LLM and GitHub API
- Map out which parts can be automated vs require direct code
- Design the GitHub Action architecture
- Outline the key steps: clone repo, get context, analyze PR diff, log comments
- Determine config options: target repo/PR, credentials, comment rules, logging
- Decide how Goose integrates into the Action flow
- Identify any dependencies and constraints
- GitHub API client library to use
- Credential management approach
- Execution environment requirements
- Configure Goose agent
- Provide prompt engineering to describe the task
- Connect to LLM and GitHub API
- Implement core logic to analyze diff and generate comments
- Develop GitHub Action shell
- Set up config parsing and validation
- Invoke Goose to perform the automated steps
- Handle cloning the repo and pushing comments
- Add support for multiple LLM providers
- Implement OpenAI integration
- Implement Anthropic integration
- Make provider selection configurable
- Convert project to TypeScript
- Add TypeScript configuration
- Define interfaces and types
- Convert JavaScript files to TypeScript
- Update tests for TypeScript
- Test and iterate
- Trial run on sample repos and PRs
- Debug and refine Goose prompts as needed
- Verify comments are appropriate and useful
- Package GitHub Action
- Document configuration and usage
- Publish to GitHub Marketplace
- Integrate into our CI/CD pipelines
- Add as a required check on PRs
- Configure for our key repositories