What problem does it solve? Before any episode-level directing or storyboard work begins, a short-drama production needs a single, consistent inventory of every character, scene, and prop that will require reference images. Doing this manually across dozens of converted scripts leads to duplicated character names, missed recurring props, and inconsistent visual variants. ## Core Features & Use Cases - Sharded Parallel Inventory: Splits all converted episode scripts into groups of five and dispatches parallel asset-analyst sub-agents, each writing a structured chunk JSON with characters, scenes, and props. - Merge and Deduplication Rules: Consolidates chunk results into one canonical list, merging name aliases, rejecting temporary poses or expressions as character variants, and retaining props that participate in causal chains, cross-scene continuity, or repeated on-screen actions. - Finalized Deliverables: Produces a human-readable full-show asset table (全剧资产总表.md) and a machine-readable final-asset-bible.json with firstEpisode, visualKeywords, decision, and mergeNotes per entry, gated by a formal validation step. - Use Case: Given 30 converted episode scripts, run this skill to produce the definitive asset bible that downstream costume/prop prompt writing and per-episode directing will reference. ## Quick Start Ask the assistant to run the full-show asset preflight over all converted scripts in 资产/converted-scripts and produce the finalized asset table and asset bible before any directing work begins.
