What problem does it solve? Runtime sessions produce a mix of reusable lessons and project-specific noise, and without a disciplined process the wrong knowledge leaks into shared capability pools or useful lessons are lost entirely. This Skill gives the Evolver agent a structured workflow to separate the two and propose durable improvements safely. ## Core Features & Use Cases - Friction Detection: Scans session events for repeated corrections, instructive errors, unclear instructions, wasted effort, and effective patterns worth preserving. - De-contextualization and Generalization Gate: Strips paths, project names, and session IDs, then applies three questions to decide whether a lesson holds beyond the current space. - Three-Tier Triage: Routes each candidate to space-private memory, type-specific assembly, or the public core pool, preventing contamination of shared assets. - Evolution Plan Output: Produces a structured Markdown proposal with evidence, tier justification, risks, and next steps, stopping before any implementation. - Use Case: After a session where the user corrected the same behavior twice, the Evolver loads this Skill, distills the underlying rule, classifies it as type-specific, and emits an Evolution Plan for the user to approve before Fae implements it. ## Quick Start Ask the Evolver agent to distill this session's lessons and corrections into an Evolution Plan for your review.
