meta-prompter

name: meta-prompter description: Automatic prompt engineering & optimization keywords: - prompt-engineering - optimization - meta-prompting - llm - tuning measurable_outcome: Improves prompt performance metrics by >15% over baseline. license: MIT metadata: author: Biomedical OS Team version: "1.0.0" compatibility: - system: Python 3.10+ allowed-tools: - run_shell_command - read_file

Meta-Prompter

The Meta-Prompter is a tool for self-optimizing agent prompts. It analyzes agent performance and iteratively refines system prompts to maximize accuracy and adherence to instructions.

When to Use This Skill

• When an agent is consistently failing a specific type of task.
• When deploying a new agent and needing to tune its persona.
• When A/B testing different prompting strategies.

Core Capabilities

1. Prompt Optimization: Rewriting prompts for clarity and effectiveness.
2. Performance Evaluation: Testing prompts against benchmarks.
3. Few-Shot Generation: Creating optimal examples for context.

Example Usage

User: "Optimize the Clinical Reasoning prompt."

Agent Action:

python3 platform/optimizer/meta_prompter.py --target "clinical_reasoning" --iterations 5