Files
codex-mmo/evals/adaptive-change/README.md
T
2026-08-24 08:11:59 -07:00

541 B

Adaptive change evaluation

This suite exercises the adaptive-engineering profile on a bounded but nontrivial correctness change. The root must keep ownership of diagnosis and integration, use a specialist only where it adds value, and obtain independent review before claiming success.

The native scout is intentionally not a worker assertion: native Codex participation is not represented in Agent MCP job telemetry. The evaluator does require the MCP implementation and review roles and measures whether their results were consumed.