Each tool has a fairly distinct purpose: audit_codebase runs the full audit, verify_changes compares fingerprints against a baseline, explain_finding drills into one finding, and describe_capabilities is meta. The only mild overlap is that audit_codebase and verify_changes both execute a read-only audit, but their descriptions make the baseline-vs-verification distinction clear.

All four tools follow a clean verb_noun snake_case pattern (audit_codebase, describe_capabilities, verify_changes, explain_finding), with no mixed conventions or vague verbs.

Four tools is a tightly scoped, coherent set for a read-only audit workflow (run, explain, verify, describe). It is on the lean side but each tool earns its place, so it lands slightly below ideal breadth rather than being bloated.

The surface covers the core audit lifecycle: run an audit, inspect individual findings, verify repairs, and introspect the server's capabilities. A dedicated findings-listing tool is absent, but findings are surfaced through the report and explain_finding, so agents can work around it; read-only design legitimately omits write operations.