camera-multiviewlisted

# camera-multiview Execute the single `multiview` stage with the bundled API workflow and the bundled pose assets. This workflow consumes only the two reference images. It has no prompt input, so the MCP envelope must be `{}` and no prompt authoring skill is involved. ## Contract Expose exactly two image inputs: - `full_body_image`: local full-body image; patch only node `111.inputs.image`. - `face_image`: local face image; patch only node `667.inputs.image`. Keep every other workflow input fixed. In particular, do not expose or infer group switches, prompts, samplers, LoRA, ControlNet, dimensions, seeds, or alternate workflows. Keep the API asset and its manifest authoritative: `camera_multiview/runtime/workflow_assets/Flux2-Klein人物一键多视图工作流.json` Hydrate `姿势骨架1.png` through `姿势骨架13.png` from the adjacent `pose/` directory. Preserve this node mapping exactly: `152,154,360,364,148,149,147,373,150,367,368,151,757` correspond to poses `1` through `13` respectively. ## Procedure 1. Pass the project preflight gate before production execution. 2. Use `list_skills`, `describe_config`, and `validate_config` for the `camera-multiview` / `multiview` contract. 3. Require both local image paths. Do not accept a partial request. 4. Load the bundled API graph and verify the manifest, node topology, titles, node count, and fixed asset hashes. 5. Reuse an existing ComfyUI input when a fixed pose filename is already present; upload it only when absent. 6. Deep-copy the g