Craft, rewrite, diagnose, compare, and package efficient prompts for generative video models and conditioning modes. Use for MiniMax H3, LTX-2.3, LTX-2.5, Wan 2.2, or another video model; for T2V, I2V, first-frame, first-and-last-frame, last-frame, reference-to-video, image/audio-to-video, speech-to-video, video-to-video, retake, extend, reframe, character animation/replacement, inpainting, multi-shot, or runtime-specific modes; for prompt-adherence, dialogue-speaker, camera, action, continuity, timing, reference-assignment, audio, or negative-prompt problems; and for converting this expertise into a system prompt, assistant specification, knowledge base, or skill for another LLM. Exclude workflow installation, hardware, quantisation, sampling speed, LoRAs used only for acceleration, and unrelated node-graph optimisation.