What problem does it solve? Changes to the pizza spec import pipeline (AI parse prompts, sanitizers, alias logic, chunk merging) can silently corrupt production data through stale parse caches, dropped rows, poisoned aliases, or mis-routed recipes. This Skill provides a mandatory validation checklist that catches these regressions before they reach managers re-importing spec sheets. ## Core Features & Use Cases - Cache Version Enforcement: Ensures SPEC_PARSE_VERSION is bumped whenever the parse prompt, sanitizer output, or pipeline ordering changes, preventing stale cached parses from being served. - Deterministic Regression Testing: Runs the corpus harness against checked-in snapshots and the lib/spec-import unit tests to catch dropped rows, routing errors, and alias hygiene violations without AI calls. - Real-AI Round-Trip Harnesses: Verifies large-spec imports and parse-rule stress scenarios against the live model to detect silent data loss when prompts or models change. - Use Case: After rewording the AI parse prompt, follow the checklist to bump SPEC_PARSE_VERSION, run the corpus tests, and execute the smoke round-trip harness to confirm no profiles or recipes are lost. ## Quick Start Ask the AI to run the spec import guard checklist for your changes to the parse prompt in aiParseSpecSheet.ts.
