Scored across 39 tools
Each tool is namespaced by domain (workspace, participant, task, review, deliberate, audit) and targets a distinct action; even the several state-changing task tools are separated by explicit verbs like complete, update, cancel, and supersede. The few routing/decision tools (route, depth, auto) have clearly different outputs despite all recording route_decision artefacts.
All tool names follow the same chap.<domain>.<verb> pattern with consistent snake_case and verb-final naming. Verbs are descriptive and domain prefixes make the action hierarchy predictable, so an agent can infer where a tool belongs.
At 39 tools, the surface is well beyond the 16-25 heavy range and even past the 25+ threshold for too many. Although the broad coordination/audit domain justifies many operations, the sheer number will make tool selection costlier and suggests some consolidation could be considered.
The surface covers nearly the full lifecycle: workspace creation/describing, participant management, task creation/state transitions, review, escalation, deliberation, handoff, control/snapshot, and SCITT audit verification. Minor gaps exist (no direct task/workspace listing or deletion), but agents can work around them via describe and audit.read.