Skip to content

Agent Directory: A collection of SFAs (Single-File Agents) using Astral UV in Python Armed with OpenAI Function Calling Tools

Notifications You must be signed in to change notification settings

seanivore/claude-agents

Folders and files

NameName
Last commit message
Last commit date
Mar 18, 2025
Mar 23, 2025
Mar 23, 2025
Mar 22, 2025
Mar 22, 2025
Mar 23, 2025
Mar 23, 2025
Mar 23, 2025
Mar 18, 2025
Mar 18, 2025
Mar 18, 2025
Mar 22, 2025
Mar 22, 2025
Mar 22, 2025
Mar 18, 2025
Mar 23, 2025
Mar 22, 2025

Repository files navigation

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