What problem does it solve? Pattern-based CI gates fail silently in two directions: they miss alternate spellings of the construct they police (bracket access, quoted keys, namespace imports), or they count commented-out and quoted code as real coverage, letting a gate pass while enforcing nothing. ## Core Features & Use Cases - AST-first matching guidance: Shows how to walk the TypeScript AST with a pure (fileName, sourceText) parser so comments and string literals never count as executable calls. - Mandatory fail-closed fixtures: Defines the three test cases every guard needs — a real call, a commented-out call, and a quoted call — plus a fourth assertion that the walk found enough calls to avoid vacuous passes. - Spelling enumeration for regex gates: Checklists recurring miss patterns across member access, object keys, imports, and function declarations, with must-fail fixtures per form. - Use Case: A drift guard greps source for a helper call and has never failed since it was added; use this Skill to prove it can fail, add fixtures for commented-out and quoted variants, and widen the pattern to cover bracket-access spellings. ## Quick Start Ask the AI to verify whether the gate script under scripts/ci/ can actually fail by adding must-fail fixtures for commented-out, quoted, and bracket-access spellings of the call it polices.
