AssurancesLabs

Each tool targets a different artifact: a sources map, a documentary leak/repair search, and a quote-intent-to-deeplink resolver. The two 'get_' tools are reads but operate on clearly separate domains, so misselection risk is low; only the shared read/retrieval framing causes slight overlap.

All names are snake_case and verb-first (get_*, resolve_*), which is predictable. Minor deviation: 'get_habitation_leak_search' embeds a noun phrase suggesting a search operation under a 'get' verb, and 'resolve_' is a slightly different verb class than the rest.

Three tools is thin for an insurance-domain server, though each maps to a distinct documented API endpoint. It reads more like a minimal AI/citation facade than a full toolset, sitting at the borderline of under-scoped.

The surface exposes read-only lookups and intent resolution but no way to fetch evidence or document details by the fact/evidence IDs it returns, and no quote submission or coverage lookup. Agents hitting the abstention path or needing to verify a citation have no follow-up tool and will dead-end.