Loading ToolBench
Institutional memory system for multiple AI agents with P2P sync, cryptographic signing, and governance. Provides a journaled, distributed corpus with device identity, owner governance, and read-authentication.
Hive Mind exhibits severe definition quality gaps across all dimensions. Tool descriptions are present but extremely terse (avg 60 chars, well below the 194-char baseline). Most critically, input schemas are either missing or severely incomplete: 15 of 20 tools lack visible parameter type definitions in the provided source. Parameter descriptions are sparse or absent. No output schemas are documented. Error handling is absent, no recovery guidance, no categorization, no actionable error messages. The server appears to expose low-level sync/API endpoints (sync_hello, sync_merkle_root, sync_chunk, sync_ingest) directly as tools, violating the single-responsibility principle. Tool names are inconsistent (api_* prefix for dashboard endpoints, sync_* for replication, hive_* for memory) and some are ambiguous (api_overview, api_status). No evidence of idempotency guarantees, pagination support, or composition chains. Security-critical tools like sync_ingest and nudge lack permission declarations or audit guidance.
Dashboard API endpoint for audit trail
Dashboard API endpoint for retrieving a single item
Dashboard API endpoint for corpus overview
Dashboard API endpoint for peer information
Dashboard API endpoint for finding related entries
Dashboard API endpoint for corpus search
Dashboard API endpoint for system status (hv whoami, stats, doctor)
15 of 20 tools have no visible input schema or severely incomplete schemas (missing type definitions).
Tool descriptions are extremely terse (avg 60 chars vs 194-char baseline). Most lack context for LLM selection. E.g., 'Dashboard API endpoint for corpus overview' does not explain when to call it or what it returns.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. E.g., does sync_hello return node_id as string or object? What fields does api_search return?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
Dashboard API endpoint for tag listing
Dashboard API endpoint for telemetry data
Open discovery endpoint for signature verification
Retrieve audit trail with optional depth and format filtering
Open discovery endpoint: returns hive id, owner id, genesis, node id, root hash, and verdict
Store memory via hive CLI
Search hive memory
Record a session event (start/end) into the LOCAL telemetry lane
Retrieve journal entries for a specific node and sequence range
Node discovery and journal summary endpoint. Returns node_id, journal summary (by_node maxima), and per-node chunk hashes.
Append foreign entries to the journal (G-Set dedup), rebuild, and return acceptance summary
Global root hash for O(1) identity check between nodes
Record a telemetry event into the LOCAL telemetry lane
Destructive/sensitive tools (sync_ingest, nudge, telemetry_record) lack permission declarations, audit guidance, or confirmation patterns. No indication of what permissions are required or how to safely invoke them.
Inconsistent naming convention: api_* (dashboard), sync_* (replication), hive_* (memory). Generic names like api_overview, api_status are ambiguous. No verb-noun pattern (e.g., get_overview, list_peers).