Scored across 40 tools
Most tools have explicit 'use X instead of Y' disambiguation, but there are multiple overlapping clusters: report detail/evidence/timeline/fix context and graph traversal/neighborhood/blast radius. An agent could easily misselect get_knowledge_graph vs get_graph_neighborhood vs get_blast_radius without close reading.
The vast majority of tools follow a clear snake_case verb_noun pattern (get_, list_, add_, remove_, search_, dispatch_, merge_). A few names like activation_status, get_two_way_comms_health, and suggest_fix deviate slightly, but the overall convention is predictable and readable.
40 tools is well above the 25+ threshold and includes many near-variant report, graph, and fix-context lookups that could be consolidated. Even though the domain is broad, the sheer number makes the set feel heavy and harder for an agent to navigate.
The core report → triage → fix → merge → verify lifecycle is well covered, with additional setup, health, lessons, docs, and BYOK tooling. However, the set references tools not actually exposed (e.g. generate_tdd_from_story) and lacks some obvious QA-story or report-management operations, leaving notable gaps.