qa-engineer

Automated verification sections are added by feeding a task file path to the QA engine, which inserts LLM-as-Judge blocks and weighted rubrics into each step. This ensures objective evaluation and high-quality delivery across implementation tasks.

LLM-as-Judge verification is an automated evaluation mechanism that applies risk-aware rubrics and thresholds to implementation artifacts. It generates structured criteria and reference patterns to objectively assess quality and traceability across task steps.

Risk-aware rubrics are generated by analyzing artifact types and risk levels within task files. The engine creates weighted criteria tailored to specific deliverables, producing structured thresholds and reference patterns for objective quality assessment.

Yes, automated rubrics ensure traceability by producing a verification summary and scratchpad analysis for each step. This auditability feature tracks implementation quality and provides objective evaluation aligned with risk-based guidelines.

Automated task verification operates independently without external dependencies. The engine processes task files directly, inserting verification sections, designing weighted rubrics, and generating audit summaries within a self-contained workflow.

Task files should be structured as step-by-step implementation plans. Providing the file path to the QA engine allows it to insert verification blocks, generate risk-based rubrics, and create traceability summaries aligned with each step.