agent-output-evaluator

# Agent Output Evaluator ## Trigger Use for quality audits of generated outputs. ## Inputs - plan/task files, skill outputs, fixtures. ## Outputs - Rubric-based score and specific violations. ## Procedure 1. Validate required task fields. 2. Validate wave safety and scope constraints. 3. Detect vague tasks and oversized PR assumptions. 4. Provide fix recommendations. ## Guardrails - Keep scoring reproducible and artifact-based. ## Success criteria - Every failure includes exact path + reason.