Three tools with minimal documentation. All have basic descriptions (10-50 chars, below the 50-200 char baseline for LLM-optimized descriptions). Input schemas present but sparse: taskId and bidId are strings with minimal context; requirements is an untyped object with no field definitions. No output schemas documented. No error handling guidance. No parameter constraints (enums, ranges, formats). Tool names are verb-noun but descriptions lack WHEN/WHY context. No indication of idempotency, retry safety, or side effects despite all three being WRITE operations.

• Expand each tool description to 100-150 chars. Include: what it does, when to call it, what it returns, and any prerequisites. E.g., 'Request competitive bids from available agents for a task. Call this after creating a task to solicit proposals. Returns a list of bids with agent IDs, proposed costs, and timelines. Agents have 5 minutes to respond.'
• Define the 'requirements' object schema explicitly. Add properties: capabilities (array of strings), budget (number, min/max), deadline (ISO 8601 string), priority (enum: low|medium|high). Add descriptions for each field.
• Document output schemas for all three tools. E.g., request_bid returns {bidId, agentId, proposedCost, timeline, agentRating}. assign_task returns {assignmentId, taskId, agentId, startTime}. notify_payment returns {transactionId, status, confirmationTime}.
• Add parameter descriptions for taskId and bidId. E.g., 'taskId: The unique identifier of the task (UUID format, e.g., 550e8400-e29b-41d4-a716-446655440000)'. Specify format constraints in descriptions since JSON Schema patterns are not parsed by LLMs.
• Add error handling guidance to each tool description. E.g., 'If no bids are received within 5 minutes, call request_bid again with adjusted requirements. If assign_task fails with "agent unavailable", try the next highest-rated bid.'