What problem does it solve? Authors using AI-assisted writing tools need recommendation engines to understand the current state of their work, but computing that context synchronously during typing freezes the editor. This Skill asynchronously compiles the author's live writing session into a single Contextual Authoring Context (CAC) vector without blocking the UI. ## Core Features & Use Cases - Multi-source context fusion: Combines recent text embeddings, active scene entities from the local lore graph, structural book position, and author feedback history into one weighted vector. - Debounced background execution: Runs computation in a Web Worker with a 1000ms debounce, triggered by paragraph breaks, shortcut keys, or 5 seconds of idle cursor time. - Configurable weighting: Exposes parameters like CAC_DEBOUNCE_DELAY_MS, RECENT_WORDS_COUNT, and WEIGHT_TEXT_RECENT to tune vector composition. - Use Case: A novelist pauses mid-scene; the Skill compiles a 768-dimension CAC vector capturing the last 512 words, the active character "Kael", and 10% book completion, feeding it to local plot and character recommendation engines. ## Quick Start Compile the current writing context vector from my editor session so the local recommendation engine can suggest relevant plot and character ideas.
