spec-harness-audit

To fix spec-to-code drift, you must audit the spec harness memory bank and rules index to identify mismatches between spec definitions and actual production code, ensuring AI coding assistants follow consistent project rules instead of hallucinating outdated patterns.

AI coding assistants generate inconsistent code when the spec harness contains stale, contradictory, or redundant entries in the memory bank and rules index. Auditing these components cross-checks for misaligned definitions and eliminates outdated patterns causing the inconsistency.

Auditing a memory bank for stale rules involves cross-checking the memory bank, ai_rules, and index to identify redundant or contradictory entries. This holistic health check ensures all stored rules are consistent and up-to-date for AI-driven development workflows.

Yes, a spec harness audit applies directly to spec-driven development workflows using AI tools like Claude Code or Cursor. It ensures the AI adheres to latest project rules by maintaining rule ratcheting, verifier pass criteria, and durable learning capture.

Durable learning capture in spec harness systems is the process of promoting validated fixes into the curated rules bank. This prevents recurring issues by ensuring AI coding assistants permanently learn from corrected spec-to-code drift and updated project rules.