Scored across 17 tools
Every tool has a clearly delineated purpose, with descriptions that explicitly contrast neighboring tools (e.g., send vs. ask vs. answer vs. decide). Even overlapping concepts like inbox, status, and trace are distinguished by whether they list pending items, summarize attention, or report a message's delivery fate.
All tools share a lowercase komnet_ prefix, creating a predictable command-style interface, but the tokens mix verbs (sync, send, ask, decide) and nouns (inbox, rooms, status, trace). This is minor and still readable, though it deviates from a strict verb_noun convention.
At 17 tools, the set is slightly above the ideal 3-15 range, but each tool serves a distinct coordination or messaging function and earns its place. The count reflects a genuinely broad domain rather than redundancy.
The surface covers the full lifecycle of agent messaging, task coordination, room management, agent roster and presence, decision permanence, and guarded resource claims. Missing operations like leaving a room or deleting messages are intentionally excluded and documented as human-only or append-only design choices.
