darwinian-evolver

Automated prompt optimization uses a fitness-guided loop to mutate and evaluate prompt variants, discovering higher-quality outputs. This Skill defines an Organism, Evaluator, and Mutator framework to iteratively search for better prompts, regexes, or small code snippets.

Yes, automated mutation can evolve SQL queries and regex patterns. The framework evaluates mutated variants against a defined fitness criteria, iterating toward better solutions for data tasks, patterns, or small code blocks.

Running an AI-driven evolutionary loop requires Python 3.11+ and an API key for a chosen LLM provider. You must define a custom Problem or use a driver like parrot_openrouter.py with your OPENROUTER_API_KEY.

The Evaluator and Mutator framework improves code snippets by applying automated mutations and scoring them against a fitness function. It iterates through high-quality variants, relying on LLM providers to generate and assess improvements.

Using Jinja2 templates for evolving prompts involves defining the initial Organism structure and mutation parameters within the template. The evolutionary loop then dynamically generates and evaluates prompt variants to optimize the output.

You should avoid using a fitness-guided loop for prompt engineering when dealing with simple, static prompts that do not require iterative improvement, or when lacking an LLM API key, as the framework relies on automated evaluation and mutation.