This server exhibits significant definition quality gaps. While 5 tools are registered with schemas and descriptions, the descriptions are highly technical and domain-specific (targeting Co-Engineer/ACP workflows), making them difficult for general LLM agents to understand. Tool names lack clear action verbs (status, delegate, task, tasks, cancel are vague). Parameter descriptions are sparse or missing context. The 'delegate' tool has a complex schema with many optional fields and interdependencies that are not documented in parameter descriptions. Output schemas are not documented. Error handling guidance is absent. The server appears designed for a specialized internal system rather than general agent composition.

• Rename tools to start with action verbs: 'get_run_status', 'start_run', 'get_task', 'list_tasks', 'cancel_run'. This enables LLMs to infer intent from names alone.
• Add 50-150 character descriptions to every parameter. For 'delegate', explain: 'provider: Which AI provider to delegate to (grok, cursor-local, cursor-cloud, dsh). Required.' For 'task', explain each of the 5 properties.
• Document output schemas in tool descriptions. E.g., 'Returns: {run_id, status, cursor, attention, tasks[]}' so LLMs know what fields to expect and can chain calls.
• Add error handling guidance. E.g., 'If delegation fails with timeout, retry with a longer expected_duration_ms. If provider is unavailable, try a different provider from the enum.'
• Clarify parameter relationships in descriptions. E.g., 'expected_duration_ms and timeout_ms are mutually exclusive. If both provided, timeout_ms must be >= ceil(expected_duration_ms * 1.20). Omit both to use 600000 ms default.'