dsh-skill-nannan

Writing a good AI video prompt is far more involved than most people expect. Character consistency, spatial relationships, performance pacing, shot-by-shot rhythm — every layer demands discipline, and scattered community tips rarely form a coherent system. dsh-skill-nannan packages the 29 imperative rules distilled from the Hell Grind production methodology into a loadable DeepSeek Harness skill, so the model can pull them up automatically whenever they are needed, without the user having to comb through long articles.

The 29 rules span four stages: asset and consistency, scene and spatial layout, character performance design, and video generation with shot-by-shot prompting. Each rule is a concrete, actionable directive rather than a vague suggestion, and key examples from the source — GEO SPATIAL LAYOUT, the VIKTOR performance profile, the corridor-scene timeline, the 30-metre stone-guard scale anchor, and style presets — are preserved so you can adapt them directly.

It is built for creators, prompt engineers, and solo developers who work with Seedance, Sora, Kling, Runway, or any other video model and need to generate or refine storyboard prompts repeatedly. The moment a conversation touches on writing video prompts, designing character performances, or converting a story into shot-level generation cues, the skill loads on its own.