What problem does it solve? Adding highlights, arrows, zoom focus, and callout cards to video footage normally requires manual motion-graphics work in an editor. This Skill turns user-selected screen regions into timed annotation definitions and renders them as transparent overlay layers composited onto the final video, with caching so unchanged annotations are not re-rendered. ## Core Features & Use Cases - Four annotation styles: highlight-rect, arrow, zoom-focus, and callout-card, defined with normalized bounding boxes and start/end times so they survive resolution changes and timeline re-edits. - Incremental rendering with caching: unchanged annotations hit the cache and are skipped, so fixing one misplaced highlight box does not trigger a full re-render. - Workbench request handling: claims pending "generate annotation layer" requests from the UI queue via request_claim and reports long render jobs through job_status. - Dependency recovery: detects a missing HyperFrames installation, quotes the download size and install location, and completes the dependency_fetch flow before rendering. - Use Case: A user框 selects a save button in a screen recording, asks to circle it from 2 to 3 seconds, and the Skill defines the annotation, renders the transparent layer, and composites it onto the footage at the exact frames. ## Quick Start Circle the save button in the current clip from second 2 to second 3 and render the annotation layer onto the video.
