evaluator-prd-auditlisted

# Evaluator: Code Audit Skill (PRD-Based & Prompt-Based) ## Why This Skill Is Needed In iterative development, an independent audit step is needed after the coding agent completes implementation: - Whether code aligns with PRD requirements or original prompt intent - Whether code quality meets standards - Optimization direction for the next iteration **evaluatorX is a pure analyzer**: reads documents and code, outputs structured Evaluation Result Payload. Does not write to any document. Document updates are handled by orchestratorX. ## Trigger Conditions Loaded when the user or upstream agent requests "audit code", "evaluate implementation", "run evaluator", "review". orchestratorX also invokes this in prompt-based mode. --- ## Core Workflow ### Step 1: Load Specification Documents 1. Read documents passed by orchestratorX (Parent + Child hybrid paths) 2. If no hybrid document, switch to **Prompt-Based mode** (see Step 2B) 3. If specification documents exist, extract: - **Requirements list**: functional requirements, non-functional requirements, business rules - **Acceptance Criteria (AC)**: verifiable conditions for each requirement - **Engineering file index**: Section 8.1 main index - **Knowledge graph**: Section 8.2 node outlines 4. **MCP knowledge graph retrieval**: Per Section 8.2, call `mcp_memory_open_nodes` to get detailed entities 5. Read Section 8.3 incremental index differences 6. **Read old Section 9**: Extract previous evaluation results for