GEP Gene Expression Programming Agent Evolution Engine Design Template v2

You are a Genetic Expression Programming (GEP) architect for AI Agents. Design a self-evolution engine that uses GEP principles to optimize an agent's tool-calling strategies and decision paths. Agent Context: - Available tools: [USER LISTS TOOLS] - Primary task domain: [USER PROVIDES] - Success metric: [USER DEFINES] Design these components: 1. Genome Encoding - Gene alphabet (tool calls as functions, parameters as terminals), head/tail length, multi-genic chromosome design, linking functions 2. Fitness Function: fitness(chromosome) = a * task_success_rate + b * (1/token_consumption) + c * (1/latency) - d * error_penalty Define weights based on user priorities. Include novelty bonus. 3. Genetic Operators - Mutation (tool substitution, parameter perturbation, gene transposition), Crossover (one-point, two-point, gene-level), Selection (tournament with elitism) 4. Evolution Loop - Initialize 50 random tool-call sequences, execute as agent workflows, evaluate fitness, select/crossover/mutate, log best to skill library 5. Skill Crystallization - When a chromosome scores above threshold for 5 generations, crystallize into a named reusable skill with documented I/O contract Output a complete implementation plan with code structure, recommended libraries (DEAP or custom), and a concrete example using the user's tools.