What problem does it solve? TypeScript code often looks type-safe while hiding real soundness gaps: unsafe as casts, any leakage, partial validation treated as full trust, utility types that collapse unions, and exported types that overpromise. This Skill performs a findings-first review that separates genuine unsoundness from missing proof, residual risk, and style-only concerns. ## Core Features & Use Cases - Soundness-Focused Review: Identifies the exact safety claim code makes, traces where compiler proof ends or runtime truth disagrees, and classifies each point as a finding, missing proof, or residual risk. - Structured Inspection Workflow: Provides checklists covering compiler strictness settings, boundary trust sweeps, escape hatches, helper composition, and public API surfaces. - Finding Calibration: Challenges the strongest nearby non-finding story before keeping a point, and recommends the smallest safe fix rather than broad rewrites. - Use Case: When reviewing a TypeScript PR that adds a parser and new exported types, use this Skill to verify whether unknown truly stops at the validation boundary and whether the exported signatures promise more than the runtime guarantees. ## Quick Start Ask the AI to review your TypeScript diff or PR for type-safety issues such as unsafe casts, any leakage, partial validation, or exported types that overpromise guarantees.
