sadd:judge

To automate code review, this Skill extracts concise work context from artifacts like code changes or documentation edits, then generates tailored rubrics and applies evidence-based scoring with cited file paths and line numbers to ensure objective, reproducible judgments.

Two-stage evaluation for conversational artifacts is a coordinated pipeline where a meta-judge first generates tailored YAML rubrics, then an isolated judge applies those specifications to produce structured scores with cited evidence, ensuring objective and reproducible assessments.

You evaluate recent code changes by dispatching an isolated judge agent with fresh context to apply generated YAML rubrics, producing structured scores and citing exact evidence such as file paths and line numbers for correctness and test coverage.

Yes, this rubric generation and scoring pipeline supports configuration edits. It applies to conversational artifacts including configuration updates, documentation, and design proposals where objective, context-isolated assessments are required.

Yes, evidence-based evaluation works without modifying the original code. It dispatches an isolated judge with fresh context to generate structured YAML reports and cited evidence, ensuring the original content remains unaltered during the assessment.

Automated rubric-based judging requires extracting concise work context and dispatching isolated agents, which may limit its suitability for extremely large artifacts. It validates score ranges, justification consistency, and weighted totals before presenting a full report.