Agentisan has 6 tools with basic schemas and descriptions, but significant gaps prevent higher scores. Tool names lack clear action verbs (inspect, team_action are vague). Descriptions exist but are brief (50-100 chars) and lack WHEN/WHY context. Input schemas use generic 'object' types with minimal field-level documentation. No output schemas documented. Error handling is server-side (HTTP status codes) but not reflected in tool definitions. Parameters lack enums, ranges, or format constraints. No idempotency guidance in descriptions despite idempotency_key parameters present in 3 tools.

• Rename 'inspect' to 'query_runs' or 'get_run_state' and 'team_action' to separate tools: 'send_message', 'create_assignment', 'request_decision'. Each should have one clear responsibility.
• Expand tool descriptions to 100-200 chars, explicitly stating WHAT the tool does, WHEN to use it (vs similar tools), and what it returns. E.g., 'Query authoritative Agentisan state for a specific run. Returns run metadata, agent assignments, and decision status. Use this to check progress before sending messages.'
• Document output schemas for all tools. Specify which fields are always present (run_id, assignment_id) vs optional, and their types. This enables LLMs to chain tools correctly.
• Convert state enum values and deadline_seconds range into JSON Schema constraints (enum, minimum, maximum) rather than prose descriptions. LLMs parse schema constraints more reliably.
• Add explicit idempotency guidance: 'Idempotency key ensures this operation is safe to retry. If the same key is reused, the server returns the original result without side effects.' This clarifies retry semantics for agents.