Create a flat competency-review table

Converts a generated benchmark into the human-review surface used by write_agentic_routing_review(). Reviewers may edit review_status, review_note, prompt, and canonical_answer. All other columns are locked provenance and are verified when the review is applied.

Usage

as_agentic_routing_review(benchmark)

Arguments

benchmark

A generated agentic_routing_benchmark.

Value

A base R data frame with one row per competency question.