passive-agent-kahiroka 0.1.1

A passive agent for prompt engineering experiments with instruction-based context building

passive-agent

A passive agent for prompt engineering experiments with instruction-based context building.

Overview

Passive Agent is a flexible tool for building complex contexts by processing instructions that combine file content with AI completions. It's designed for prompt engineering experiments where you need to iteratively build context from multiple sources.

Features

• 📄 Load content from files and directories
• 🤖 Integrate AI completions into context (supports OpenAI and OpenRouter)
• 🔄 Sequential processing with context accumulation
• 📊 Token usage tracking and reporting
• 🎯 Simple instruction-based workflow

Installation

pip install passive-agent

Or install from source:

git clone https://github.com/yourusername/passive-agent
cd passive-agent
pip install -e .

Quick Start

1. Create an INSTRUCT.md file with your instructions:

   @header.md
   @data/
   /completion
   @footer.md
   

2. Run passive-agent:

   passive-agent
   

3. Check the generated files:

   • CONTEXT.md - Complete context with all content
   • COMPLETION.json - Raw API response
   • COMPLETION.md - Completion text

Example

See the example/ directory for a complete working example:

cd example
passive-agent

Configuration

OpenAI

export OPENAI_API_KEY="your-key"
passive-agent

OpenRouter

export OPENROUTER_API_KEY="your-key"
export OPENROUTER_MODEL="anthropic/claude-3-opus"
passive-agent

Documentation

For detailed usage instructions, see USAGE.md.

License

MIT License - see LICENSE file for details.