Claude Agents

Overview

A collection of modular, variable Single-File Agents (vSFAs) built on Claude's API for automated research, protocol development, and specialized task execution.

What are Variable Single-File Agents?

Variable Single-File Agents (vSFAs) are modular, configurable Python scripts that:

• Include all dependencies directly in the file
• Accept standardized variables to define their behavior
• Can be reused across multiple use cases by changing variables
• Use Claude's API for reasoning and decision-making
• Run with minimal setup requirements

For more detailed information on the vSFA pattern, see VARIABLE_MODULAR_SFAS.md.

Quick Start

1. Ensure you have Python 3.9+ and uv installed
2. Copy .example.env to .env and add your Anthropic API key
3. Set up the directory structure with ./setup-vsfa.sh
4. Try a predefined use case: job-search

Available Agents

The project includes both traditional Single-File Agents (SFAs) and variable Single-File Agents (vSFAs):

SFAs

Single-purpose agents for specific tasks:

• best_docs_gap_finder: Identifies gaps in documentation
• best_docs_guide_creator: Creates documentation best practice guides

vSFAs

Modular agents that can be customized with variables:

• protocol-agent: Creates detailed protocol documents from existing material
• research-agent: Conducts research based on protocol documents

Use Cases

Predefined use cases combine vSFAs for complex workflows:

• job-search: Analyzes candidate background and searches for matching job opportunities

Technical Specifications

For detailed technical information about the architecture, components, and implementation patterns, see SPECIFICATIONS.md.

Development Notes

For Claude-to-Claude collaboration notes and development insights, see CLAUDE.md.

Features

• Variable-based design: Define different behaviors with standardized variables
• Agent chaining: Connect multiple agents in sequence for complex workflows
• Self-contained files: Each agent operates as a single file with embedded dependencies
• Minimal footprint: Focused, purpose-built agents that do one thing well
• Modern dependency management: Uses uv for fast, reliable execution
• Custom commands: Create executable shortcuts for specific use cases

Development Setup

• Required: Python 3.9+, uv package manager
• API keys: Anthropic API key (Claude 3.7 Sonnet or higher recommended)
• Configure environment variables in .env file

Creating New Use Cases

1. Identify the general task and necessary variables
2. Create JSON configuration file(s) for the vSFAs
3. Create a shell script to chain the agents if needed
4. Add to ~/bin for easy command-line access

License

MIT License