Scriptwriter Master Agent

Problem

Video production often has a gap between a rough creative prompt and storyboard-ready structure. A user may provide only “make a coffee video” or a partial draft; if passed directly to a downstream storyboard agent, the result may miss characters, scenes, tone, and emotional pacing, or expose raw JSON and shot terminology too early. This skill isolates scriptwriting as a front-end step: it outputs readable Chinese Markdown scripts for users and generates structured data only after confirmation.

How It Works

• Mode detection: short or vague inputs trigger divergent expansion; fuller drafts trigger supplementary expansion, preserving the original core story without adding or removing key elements.
• Information collection: checks video type, duration, audience, tone, key messages, subject, and style; asks concrete option-based questions when critical fields are missing.
• Script drafting: produces project setup, cast and props, scene narrative, key visuals, dialogue or narration, and pacing, using concrete visual language without close-up, camera move, or timecodes.
• Confirmation and handoff: after user approval, silently creates JSON aligned with storyboard-master, such as narrative_text, key_visual_moments, and emotional_tone, then invokes the storyboard agent.

Scope

Best for short videos, ads, brand films, product demos, and similar content-creation workflows at the script stage. It does not plan shot sizes, camera movement, cuts, or transitions. If the user asks for a storyboard first, the script confirmation step should still be completed before handoff. Revisions should be full rewrites, not partial diffs, so users can read the complete version.