ai-asset-generationlisted

# AI Asset Generation (produce one asset — the call + save layer) This skill is the **mechanics** layer: how to actually call a generation model and save the result. It does NOT own the video WORKFLOW (which asset to make when, keyframe→clip sequencing) — that is the **`using-storyboard`** skill. It also does not own the deep craft — those live in focused skills the orchestration pulls: - **`realistic-image-generation`** — how to make a good realistic image / keyframe (gpt-image-2 picker, anti-AI-look tokens, selfie/demographic templates, anatomy plausibility + validation). - **`physical-action-video`** — how to make a hard physical-manipulation beat survive (FLF-first, prompt decomposition, model-escalation ladder, editorial fallback). What stays HERE: provider/model/cost mechanics, prompt building, run/poll, import + provenance, and the two **universal video invariants** (no in-video text; native audio on — Step 6.6). > **Related:** If the request features a recurring character or item — a named person, a specific product, a recurring prop, regardless of video type — check the `using-character-library` skill first and proactively surface any catalog match. The cross-piece catalog may already have a representative image you can use as a reference for image-to-image or character-consistent generation, saving a fresh generation entirely. Use this skill any time the user asks to generate visual, audio, or 3D content with AI. Do not call generation-MCP tools directly —