ui-to-codelisted

# UI to code Implement a visual design in the user's actual stack while preserving an honest boundary around reviewer independence. ## Current availability boundary The coordinator has no image or binary attachment contract. Managed cross-family structural extraction is **temporarily unavailable**. Do not pass paths, base64 image bytes, invented media fields, or raw provider commands. When the user requires an independent visual extraction as a governance gate, return typed unavailable and stop that gate. ## Workflow 1. Inspect the image using the active host's native visual capability. When any region is detail-bound — small text, dense charts, closely spaced lines, fine UI detail — magnify it before reading: use the host's native zoom or crop actions, or a code-execution crop from the full-resolution source. A full-view-only read of sub-patch detail is unreliable and tends to fail confidently rather than uncertainly. 2. State that the structural read is primary-only; do not call it a two-read or cross-family result. 3. Extract a working spec: layout regions, elements, visible text, hierarchy, approximate spacing and colors, interaction hints, responsive behavior, and accessibility requirements. 4. Inspect the project before writing. Reuse its framework, components, design tokens, naming conventions, tests, and accessibility patterns. 5. Implement in the trusted primary's normal mutation boundary. This public skill does not grant a provider