The three tools each target a clearly distinct action: creating new images, editing existing ones, and reading model metadata. An agent can trivially choose between generate_image and edit_image based on whether it has source images, and list_models is purely informational.

All names follow a consistent snake_case verb_noun pattern (generate_image, edit_image, list_models), with plain, predictable verbs and no style mixing.

Three tools is on the lean side but covers the core generate/edit/metadata loop without redundancy. A slightly larger surface (e.g. mask creation, result retrieval) would not feel padded.

The generate/edit/list lifecycle is covered and the descriptions clarify how results land on disk, but there is no way to fetch or clean up generated files, no variation/upscale path, and no way to inspect saved outputs after the fact.