Define an agentic-routing competency question — new_agentic_routing

Creates a human-reviewable competency question and its scoring rubric. The rubric is not included in the prompt sent to Codex.

Usage

new_agentic_routing_question(
  id,
  prompt,
  required_paths,
  allowed_paths = character(),
  expected_terms = character(),
  forbidden_terms = character(),
  weight = 1,
  canonical_answer = NULL,
  review_status = "approved",
  artifact_type = NULL,
  source_heading = NULL,
  source_hash = NULL,
  derivation_template = NULL,
  review_note = NULL
)

Arguments

id: Stable question identifier.
prompt: Task prompt presented to Codex.
required_paths: Repository-relative evidence paths that a successful response must cite. A path ending in / matches any descendant.
allowed_paths: Additional relevant evidence paths that may be cited without reducing routing precision.
expected_terms: Case-insensitive literal terms expected in the answer.
forbidden_terms: Case-insensitive literal terms that indicate an incorrect or superseded answer.
weight: Positive weight used in the aggregate health score.
canonical_answer: Optional maintained answer text used for deterministic token-level grounding scores.
review_status: Whether the question is "approved", "pending", or "rejected" by human review.
artifact_type: Optional generated artifact classification.
source_heading: Optional generated source-section heading.
source_hash: Optional hash of the source file used for derivation.
derivation_template: Optional identifier of the deterministic template that generated the question.
review_note: Optional human-review rationale or editing note.

Value

An object of class agentic_routing_question.