Loading ToolBench
MCP server that turns Claude Code into a team lead coordinating Claude Code / Codex / Gemini worker agents, with true-push worker→lead notifications via Claude Code's FileChanged + asyncRewake hook
15 tools with basic verb_noun naming (team_create, member_add, etc.) and present descriptions, but significant gaps in schema completeness and parameter documentation. All tools have descriptions (10-50 chars, below the 50-200 char LLM-optimized baseline), but most lack detailed parameter constraints, enums, and output schema documentation. No tool annotations (readOnlyHint, destructiveHint) despite clear risk classifications. Error handling is minimal, no recovery guidance or actionable error messages visible in source. Schemas present but sparse: most parameters lack type constraints, ranges, or format specifications. Composition is reasonable (single responsibility per tool), but output schemas are not documented for LLM planning.
Acknowledge and mark inbox messages as read
List pending messages in the lead's inbox
Add a worker member to a team
List all members in a team
Remove a member from a team
List messages in a room or conversation
Send a message to a room or member
Create a communication room in a team
List all rooms in a team
Descriptions are too short (10-50 chars vs. LLM-optimized 50-200 char baseline). Examples: 'Create a new agent team with a given name' (43 chars), 'List all teams for the current project' (38 chars). LLMs cannot determine when/why to select tools without richer context.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite explicit risk classifications in metadata (WRITE, DESTRUCTIVE, READ_ONLY). Agents cannot infer safety/retry semantics without these hints.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | <=2025-11-25 | v2 |
Create a new agent team with a given name
Delete a team and shut down its workers
List all teams for the current project
List all workers in a team
Spawn a new worker agent with a given backend
Stop a running worker
Parameter descriptions are minimal or missing. 'backend' param in member_add has description 'Backend type (claude_code, codex, gemini)' but no enum constraint in schema. 'mentions' in message_send is 'Optional mentions array' with no item type or structure documented.
Output schemas not documented. Tools like team_list, member_list, room_list, message_list, worker_list, inbox_list return lists but no schema visible for list item structure, pagination, or total count. LLMs cannot plan downstream calls without knowing response structure.
No error handling guidance visible. Destructive tools (team_delete, member_remove, worker_stop) lack confirmation/dry-run patterns or recovery instructions. No actionable error messages (e.g., 'Team not found. Available teams: [...]').