Loading ToolBench
Dispatches multi-agent work streams for marketing and content production.
This MCP server has a single tool 'dispatch_task' with critical quality gaps. The tool lacks proper parameter descriptions, has no documented output schema, provides no error handling guidance, and the implementation is hardcoded logic rather than a genuine integration. The source shows no MCP server implementation, only a Python client class and usage example. No evidence of MCP protocol registration, no tool schema in JSON Schema format, and no transport mechanism.
Dispatches a goal description across virtual team members, generating a multi-phase execution plan. Analyzes input to route tasks to appropriate specialist agents (e.g., copywriting, SEO auditing, business intelligence).
No input schema visible. The tool definition shows only {'goal_description': {'type': 'string', 'description': '...'}} but this is informal. No formal JSON Schema with required fields, minLength, maxLength, or pattern constraints.
Tool description is 151 characters, which is acceptable, but it does not explain WHEN to use this tool vs alternatives, what the output structure is, or what fields the agent should expect. Description lacks the prompt-engineering rigor required for LLM selection.
Parameter 'goal_description' has a 66-character description ('The task or goal description to be dispatched across the team') but does not specify format constraints, length limits, or examples of valid inputs. No guidance on whether the agent should pre-process the goal or call a discovery tool first.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 19 | 2026-07-28+ | v2 |
No documented output schema. The code shows dispatch_task returns {'dispatch_plan': [...], 'agent_count': ...} but there is no formal description of the dispatch_plan structure, the fields in each phase object, or whether the response is paginated. LLM must reverse-engineer from the implementation.
No error handling or recovery guidance. The implementation uses simple keyword matching (if 'copy' in desc...) with no validation, no bounds checking, and no error messages. If the goal_description is empty, malformed, or very long, the tool silently returns a default plan with no feedback to the agent.
No MCP server implementation detected. The repo contains only a Python client class (TeamCollaborationDispatcherClient) and example usage. There is no evidence of MCP transport, tool registration via MCP protocol, or communication mechanism. This is not a valid MCP server.
Hardcoded dispatch logic. The tool's behavior is entirely determined by keyword matching on 'copy', 'write', or 'marketing'. Real dispatch systems need access to a team registry, skill inventory, and workload state. This implementation cannot scale and cannot adapt to new team members or specializations.
No schema constraints on input. A well-formed schema would include minLength, maxLength, pattern, or enum constraints. The bare 'type: string' invites the LLM to pass arbitrarily long, malformed, or injection-attack inputs.