What problem does it solve? Legacy migration projects lack an owner for visual translation: architects cover structure, testers cover parity, and coders end up improvising layouts, colors, and messages. This Skill closes that gap by converting every legacy screen into an executable specification the coder can implement without inventing UI details. ## Core Features & Use Cases - Two-phase workflow with mandatory human decision: Phase 1 detects source/target platforms (COBOL TUI, Delphi VCL, VB6, WinForms, ASP, JSP, PHP, Android XML, iOS XIB) and forces an explicit choice between literal, modernized, or hybrid translation modes; Phase 2 generates the specs only after approval. - Executable screen specs: Produces target_screens.md with per-screen YAML specs in formats matched to the origin-target pair (ansi-byte-stream, component-tree, route-component, composable), plus a deviation log and optional golden files for parity testing. - Design-token enforcement: Maps every color, spacing, and typography value to design-system tokens, deriving new tokens when the legacy has no match and tracking each divergence as an approval-blocking deviation. - Use Case: Migrating a COBOL banking application to a Go CLI β the agent inventories each DISPLAY/ACCEPT screen, gets human approval for literal mode, and emits byte-exact ANSI specs with golden files so the Inspector can run constructive parity tests. ## Quick Start Ask the agent to analyze the legacy repository, detect the source UI platform, present the literal, modernized, and hybrid translation modes with trade-offs, and wait for your mode decision before generating the target screen specifications.