What problem does it solve? Writing prompts that hold character identity, camera language, audio, and scene structure across Higgsfield's video and image models is difficult, and weak prompts burn generation credits on rejected takes. This Skill routes creative requests through a dispatcher into 32 specialized sub-skills so prompts are built with tested vocabulary, model-specific constraints, and consistency anchors instead of guesswork. ## Core Features & Use Cases - Routed prompt construction: A root dispatcher directs requests to sub-skills covering Seedance, Cinema Studio, Soul character sheets, acting direction, camera controls, audio, and scene structure, applying the MCSLA prompt formula and shared negative constraints. - Consistency and failure-mode tooling: Character anchor blocks, reference-sheet conventions, staging templates, and documented failure modes (orphan limbs, fight-scene choppiness, texture drift) keep multi-shot productions coherent. - Learning memory and validation: Python scripts log every generation to a ledger, compute iterate-vs-batch verdicts, lint Seedance prompts, and validate the library before release. - Use Case: A creator planning a multi-shot UGC ad describes the product and scene; the Skill selects the right model, builds a structured prompt with camera, lighting, and audio layers, attaches character anchors, and logs the result for iteration analysis. ## Quick Start Ask the assistant to write a Higgsfield video prompt for your scene, describing the subject, setting, camera movement, and mood you want.
