A/B test agent variants for quality and token cost.
Scaffold vexjoy-agent operator .md files: frontmatter, routing block, operator context, reference loading table, phase/gate workflow.
Background memory consolidation — overnight review, merge, and injection payload for memory files.
Classify user requests and route to the correct agent + skill. Primary entry point for all delegated work.
Retrospective reader: the negative-results registry plus routing and review telemetry.
Create and iteratively improve skills through eval-driven validation.
Evaluate skills: trigger testing, A/B benchmarks, structure validation, head-to-head bake-offs.
Closed-loop toolkit self-improvement: discover gaps, diagnose, propose, critique, build, test, evolve.