capability-evolver

Auditable self-evolution works by analyzing runtime history to identify AI agent inefficiencies, then applying protocol-constrained mutations to produce reversible, traceable improvement assets. It uses Genes, Capsules, and EvolutionEvents to standardize and store these changes.

You can automate AI agent self-repair by running the continuous evolution loop via node index.js. The system scans runtime logs for errors and patterns, emitting repair-oriented directives and hypothesized changes gated by a safe, auditable protocol.

Yes, the self-improvement approach supports rollback for AI agents through deterministic, reproducible changes. It tracks mutations and personality state using memory graphs and standardized assets, ensuring any applied evolution is fully traceable and reversible.

Yes, you can configure safety checks and governance using EVOLVE_STRATEGY presets. These flexible governance configurations balance innovation and stability, ensuring protocol-constrained evolution and dynamic local tool integration remain safe.

The GEP protocol standardizes evolution prompts and stores generated Genes, Capsules, and EvolutionEvents for reuse. This ensures that AI agent improvements remain deterministic, reproducible, and fully auditable across the lifecycle.

Yes, dotenv is required as a dependency to manage environment configurations for the evolution loop. It supports the portable operations and dynamic local tool integration needed to generate auditable Gene and Capsule assets.