What problem does it solve? Turning a text prompt into an image file normally means knowing which diffusion server is running, which dialect its API speaks, and which sampling settings its checkpoint needs. This Skill routes a single apx image command to whichever engine is configured and reachable, so the caller never manages endpoints, dialects, or per-engine defaults. ## Core Features & Use Cases - Multi-engine routing: Tries engines in a configured order (AUTOMATIC1111/Forge/Draw Things, stable-diffusion.cpp, OpenAI-compatible cloud, mock) and skips unreachable servers instead of failing. - Layered option resolution: Merges global defaults, per-engine defaults, and per-call flags like size, steps, guidance, seed, and negative prompt, so checkpoint-specific settings live with the engine. - Ignored-option reporting: Reports which options an engine could not honor (e.g. seed or negative prompt on the OpenAI dialect) instead of silently dropping them. - Use Case: Ask for a logo sketch with a fixed seed and specific dimensions; the command routes to the local A1111 server, applies that engine's step and guidance defaults, writes the PNG to ~/.apx/images/, and prints the file path. ## Quick Start Ask the agent to generate an image of a green origami fox on a white background and save it to a file using the apx image command.
