This repo is a shared, centralized location for the team to put common workflows and tools that we've found useful in our AI use. While the team is the main contributors/users, external contributions are certainly welcome.
The majority of this repo focuses around Claude Code integrations. You can pick and choose which tools and features you want, but we have an ever-evolving recommended set here which can be automatically installed with ai-cookbook recommended --yes
.
- Claude Code - Commands - reusable slash commands that we use every day.
- Claude Code - Code Standards - supports multiple languages with minimal token overhead.
- Claude Code - Hooks - before/after hooks that put Claude Code on rails. e.g.
gofmt
after Claude touches a.go
file. - Scripts - scripts that we find useful, like programatically creating
CLAUDE.md
recursively in a project
-
Clone this repository:
git clone https://github.com/ethpandaops/ai-cookbook.git cd ai-cookbook
-
Run the installer to set up all tools:
./setup.sh
-
Install recommended tools (fastest option):
ai-cookbook recommended
This installs the team's recommended configuration automatically.
-
Or launch the interactive installer for custom selection:
ai-cookbook
Use arrow keys to navigate, Enter to select components, and 'q' to quit.
-
To update everything, simply run:
git pull ai-cookbook recommended
-
To uninstall all installed components:
ai-cookbook uninstall
The following commands are available after installation:
Initializes top-level AI documentation for the entire project with foundational structure and project-wide rules. Use this when setting up AI documentation for a new project to establish consistent coding standards and development workflows.
Reads and loads project context files (README.md, CLAUDE.md, docs) into Claude's working memory. Use this at the start of a Claude session to ensure Claude understands your project structure and conventions.
Initializes AI documentation for a specific component/directory within a project with component-specific rules. Use this when adding AI documentation to individual components or modules within an existing project structure.
Executes the same action across multiple repositories in parallel for batch operations. Use this when you need to analyze patterns, implement changes, or audit multiple repositories simultaneously.
Creates a detailed implementation plan with scaffolding for features or system enhancements (deprecated - use v2). Use this when you need a structured approach to implementing complex features with clear milestones.
Generates comprehensive implementation plans optimized for maximum parallelization during execution. Use this for planning complex features where you want Claude to identify and execute independent tasks concurrently.
Enhanced implementation planning with improved analysis, dependency management, and execution tracking. Use this for the most sophisticated planning needs with automated progress monitoring.
Facilitates systematic review of implementation plans with step-by-step analysis and feedback collection. Use this to review and refine implementation plans before execution, ensuring all proposed changes align with project requirements. Usually run after create-implementation-plan-v2
has generated the plan.
Analyzes Ethereum Improvement Proposals by fetching content and optionally deep-diving into implementations. Use this when you need to understand an EIP's specifications and see how various clients have implemented it.
Creates a temporary feedback loop script that iterates with Claude until a success condition is met. Use this when you need Claude to repeatedly attempt a task (like fixing tests or meeting performance targets) until it succeeds.
Creates succinct and effective presentations in Marp format with automatic HTML generation. Use this when you need to generate professional presentations from complex topics with clear, space-efficient slides.
Enables one-shot implementation mode for automated end-to-end feature development with issue creation, parallel implementation, and PR creation. Use this when you want Claude to autonomously implement a complete feature from planning through CI monitoring without manual intervention.
Individual Claude Code command templates that can be installed/uninstalled separately:
- Each command is a standalone
.md
file with specific instructions - Install individual commands or all commands at once
- Commands are immediately available in Claude Code after installation
Language-specific coding standards that can be managed individually:
- Go:
go/CLAUDE.md
- Go formatting, naming conventions, error handling - Python:
python/CLAUDE.md
- PEP compliance, type hints, documentation - Rust:
rust/CLAUDE.md
- Memory safety, error handling, cargo conventions - TailwindCSS:
tailwindcss/CLAUDE.md
- Utility-first CSS patterns - Install/uninstall standards per language
- Automatically referenced by Claude when editing files
Individual automation hooks for different languages and tools:
- eslint: JavaScript/TypeScript formatting with ESLint
- gofmt: Go code formatting after edits
- golangci-lint: Go linting and static analysis
- typescript: TypeScript type checking after edits
- Choose global (all projects) or local (current project) installation
- Install/uninstall hooks individually based on your needs
Utility scripts added to your system PATH:
- init-ai-docs.py: Initialize AI documentation in projects
- install-hooks.py: Legacy hooks installer (still functional)
- Scripts become globally accessible commands after installation
- Operating System: macOS, Linux, or Windows (with WSL)
- Python: 3.8 or higher
- Shell: bash, zsh, or fish
- Git: For cloning and updating the repository
- Claude Code: Latest version recommended
- Claude Account: Active subscription for AI features
- Internet Connection: Required for Claude API access
- Go: Required for
gofmt
andgolangci-lint
hook functionality - Node.js/npm: Required for
eslint
andtypescript
hooks - Make: Required for some project automation
- Create a branch for your changes
- Add your commands, documentation, or scripts
- Test your changes with the installer
- Update documentation as needed
- Create a pull request for review
- After approval your changes will be merged to main