Visual Skills

What Visual Skills does

Visual Skills contains two skills. The video skill writes AI video prompts like a director, screenwriter and editor, and the image skill writes image prompts like an art director. Each one picks the right model for the task, applies its exact syntax and returns a copy-paste-ready prompt. Covered models are Seedance 2.5, Kling 3.0 Turbo and Omni, Veo 3.1, Nano Banana 2 and GPT Image 2.5.

The README puts dramaturgy before syntax. The video skill centers on a reference file, video/references/dramaturgy.md, with a scene formula of desire, obstacle, geometry, gaze and rhythm, a Details Law that gives every shot physical facts, and Walter Murch's Rule of Six for deciding where to cut. Words such as cinematic, epic, stunning and masterpiece are banned because they stand in for details the writer failed to invent. The README is available in English and Russian.

Key features

• A video skill for AI video prompts built on film dramaturgy
• An image skill for art-directed image prompts
• Model-specific syntax for Seedance 2.5, Kling 3.0, Veo 3.1, Nano Banana 2 and GPT Image 2.5
• A scene formula of desire, obstacle, geometry, gaze and rhythm
• Murch's Rule of Six used to decide where to cut
• A banned-words list that rejects empty descriptors

When to use it

• Turn a vague video idea into a shot-by-shot prompt for a chosen model
• Write an art-directed prompt for an image model
• Plan a 5 to 30 second clip as a sequence of shots with camera, light and sound

Who it is for: Creators who generate AI video and images through an agent and want prompts built on filmmaking rules instead of generic adjectives.

How it fits with Hermes Agent

It is a skill pack tagged hermes-agent alongside claude-code, cursor, opencode, windsurf and cline, so Hermes Agent is one of several agents it targets.

FAQ