pixiv_detail (metadata only) and pixiv_download (fetch files by PID) are clearly distinct, but pixiv_search already downloads originals and returns metadata, so an agent may not know whether a subsequent pixiv_download call is needed. The descriptions partly resolve this (search = discovery + fetch, download = explicit PID fetch), but the boundary between search and download remains blurred.

All names use a consistent pixiv_ prefix with snake_case and a verb/action orientation (search, download) plus one descriptive noun form (illust_detail). Minor deviation in structure but predictable and readable.

Three tools is on the thin side but each maps to a real step of the workflow: discover, inspect, fetch. No redundant or filler tools, so the small surface is justified for the server's scope.

Search, metadata inspection, and download cover the core image-retrieval lifecycle with shared filtering rules applied consistently. Gaps exist around artist/user-based browsing or paging beyond search results, but agents can work around these by re-searching with tags.