aif-evolve

To automatically improve AI skills based on project context, you can run a tool that analyzes committed patches and codebase patterns to generate project-specific prevention points and update skill-context rules. This process continually refines skill performance against your specific codebase.

Patch-driven skill evolution is a mechanism that analyzes committed code changes to extract concrete prevention points, updating skill-context policies to guard against repeating mistakes. It preserves base skill definitions while applying safe, auditable improvements tailored to your codebase.

To customize built-in AI skills for your codebase, apply project-level overrides via skill-context data. This tailors skill behavior by analyzing accumulated patches and code patterns, outputting updated rules to your skill-context directory.

Yes, AI skill evolution uses cursor-based incremental processing with patch tracking and overlap handling. This allows it to efficiently analyze only new changes and apply targeted updates to skill-context policies without reprocessing the entire project history.

A key limitation is that evolving AI skills requires access to patch directories, skill-context data, and evolution logs. The tool only outputs updated policies to specific directories, so it cannot modify base skill definitions directly and relies entirely on existing project context.