Trigger continual harness refinement from the Python REPL. Use when you notice a repeated failure, reusable tactic, delegation role, or behavior policy that should be persisted as a harness entry. Returns immediately; refinement runs when the current turn ends.
