Scored across 4 tools
The four tools split cleanly along generation (chamber_ask), self-authored claim checking (chamber_check), stored-belief drift detection (chamber_verify), and index inspection (chamber_corpus), and the descriptions explicitly cross-reference each other to steer selection. The near-synonym pair check/verify still invites a first-glance mix-up, though the read-only-vs-write and claim-vs-belief distinctions resolve it.
All four names share the chamber_ prefix and a single lowercase word, giving a predictable namespace. The pattern is verb-based for three tools (verify, check, ask) but noun-based for corpus, a minor deviation that stays readable.
Four tools is a tight, well-scoped surface for a citation-verification server, and each maps to a genuinely different operation rather than a variant of another. Nothing feels padded or redundant.
The surface covers verification, generation, and index inspection, but there is no tool to enumerate stored beliefs, retract a recorded belief, or resolve open citation debt — operations the descriptions imply exist behind the commit gate. Agents can read and verify but cannot manage or clean up the belief store they populate, which is a notable dead end.