Each tool occupies a distinct layer of a diagnosis workflow: stack_health grades, stack_snapshot returns raw data, stack_explain decodes a specific code. The descriptions explicitly state when to use each and point at the alternatives, leaving almost no room for misselection.

All three tools follow an identical `stack_<noun>` snake_case pattern with no stylistic deviation. The shared prefix also reinforces that the tools belong to one coherent family.

Three tools is on the low end, but each covers a genuinely separate need (grade, raw evidence, code explanation) with no filler or overlap. The surface is intentionally narrow rather than arbitrarily thin, though a fourth operation (e.g. history or remediation) would round it out.

The read-only diagnostic lifecycle—detect (health), inspect (snapshot), interpret (explain)—is fully covered with no dead ends within that scope. Gaps are around acting on findings: no way to acknowledge, remediate, or compare against prior snapshots, so agents can diagnose but never resolve.