Score one agentic-routing response — score_agentic_routing

Scores evidence routing, expected answer content, forbidden content, and completion using a fixed transparent rubric. This function supports offline scoring without invoking Codex.

Usage

score_agentic_routing_run(question, response, completed = TRUE)

Arguments

question: A question created by new_agentic_routing_question().
response: A list with answer, evidence_paths, route_summary, and confidence.
completed: Whether execution and structured-response parsing succeeded.

Value

A named list of component metrics and the combined score.