What problem does it solve?
Development sessions generate valuable discoveries that are often lost or siloed; agent-evolution ensures past fixes, causal learnings, and patterns are surfaced before work begins and captured after sessions so the team steadily improves its shared skill library.
Core Features & Use Cases
- Pre-session retrieval: Locate and surface past KNOWLEDGE entries relevant to a branch or feature so context is available before work starts.
- Automated knowledge capture: Read WORKLOG, CHANGES, and COMMANDS files and append concise, timestamped entries to the appropriate SKILL.md *_KNOWLEDGE sections.
- Shared-skill promotion & audits: Detect cross-agent patterns for promotion to shared skills, flag oversized SKILL.md files, and maintain an evolution log of all changes.
- Governance & traceability: Enforces entry formats, line-count thresholds, and records session summaries in references/evolution-log.md for historical review.
Quick Start
Run agent-evolution at session end to scan the session WORKLOG/CHANGES/COMMANDS, append standardized KNOWLEDGE bullets to relevant SKILL.md files, and write a summary entry to references/evolution-log.md.