What problem does it solve?
This Skill automates and optimizes the evolution of agents within the Claude Code framework, ensuring they remain efficient, relevant, and robust over time.
Core Features & Use Cases
- Automated Evolution: Tracks and evaluates agent performance, automatically proposing improvements based on session metrics.
- Data Collection & Analysis: Gathers and analyzes session data to identify areas for improvement, such as high rephrase rates, correction counts, and skill abandonment rates.
- Divergence Classification: Categorizes identified issues into types like STALE, INCOMPLETE, MISLEADING, and more, using specific criteria.
- Hypothesis Generation: Generates hypothesis based on observed signals and divergence types, detailing proposed changes and predicted outcomes.
- Surgical Edit Protocol: Proposes and implements minimal changes to agent descriptions, skill configurations, and other elements, ensuring safety and model-agnosticity.
- Prior Evolution Reconciliation: Reviews the effectiveness of past mutations, adjusting and refining the evolution protocol as needed.
- Safety Rules & Rollback Awareness: Adheres to strict safety rules and provides a rollback mechanism for undoing changes, ensuring system integrity.
- Use Case: When running agent/skill/routing/config evolution within Claude Code, load this skill to analyze session metrics and propose improvements.
Quick Start
Load the evolution skill during agent evolution to analyze session metrics and propose improvements.