Three tools with complete schemas and descriptions, but significant gaps in parameter documentation and error guidance. Tool names follow verb_noun convention (append_delta, pull_deltas, list_tasks). Descriptions are present (avg 120 chars) but lack actionable context for LLM selection. Parameters have types but most lack descriptions (only task_id, agent_id, content, kind have brief descriptions). No output schema documentation. Error handling returns raw error strings without recovery guidance. append_delta returns dict with 'error' or 'cursor' but structure is undocumented. pull_deltas and list_tasks return lists but field structure is inferred from code, not declared in schema.

• Add explicit output schema documentation to each tool. For append_delta, document: 'Returns {cursor: int} on success or {error: string} on failure.' For pull_deltas: 'Returns list of {id: int, task_id: string, agent_id: string, timestamp: string, content: string, kind: string}.'
• Enhance error messages with recovery hints. Instead of returning raw exception, return structured errors: {error: 'Missing task_id', hint: 'Provide a non-empty string for task_id', retryable: false}.
• Add enum constraint to 'kind' parameter in append_delta schema: {type: 'string', enum: ['note', 'result', 'tool_output'], description: 'Type of delta...'}.
• Expand tool descriptions to explain domain context and selection criteria. E.g., 'Pull deltas (findings/results) for a task, ordered by timestamp. Use this to retrieve updates from other agents working on the same task. Returns up to 100 deltas; use since_cursor for pagination.'
• Add limit and offset parameters to pull_deltas to enforce pagination: limit (1-100, default 20) and since_cursor (default 0). Document max result size in description.