uberskillevolver

What problem does it solve?

Prevents skills and agent workflows from drifting by capturing post-run lessons with evidence, then promoting only high-value, human-reviewed improvements.

Core Features & Use Cases

• Evidence-first learning loop: separates observations from lessons and links decisions to concrete run evidence.
• Promotion gate to avoid bloat: converts only repeated or severe patterns into durable changes (eval seeds, validators, templates, or deletion/simplification).
• Anti-regression coverage: includes structured checks for scope-fidelity failures, completion-claim regressions, red/green false-green issues, and runtime topology lessons.

Quick Start

Use uberskillevolver after a Tier 2/3 ubergoal run that changed a skill, surprised you, or failed, and produce a post-run learning record that decides what to promote, defer, or delete based on benefit >> cost.