Three tools with minimal quality. plan_swarm has a generic description (59 chars) lacking WHEN/WHY context and no visible input schema validation. get_dag_status has empty input schema but no description of what 'current swarm state' means or what fields are returned. validate_safety description (58 chars) is vague about what 'AgentFirewall checks' entails or what the output structure is. No tool has documented output schemas, error handling guidance, or parameter constraints. No tool annotations (readOnlyHint, destructiveHint). Naming is acceptable (verb_noun pattern) but descriptions fall well below the 50-200 char LLM-optimized baseline. No evidence of idempotency, pagination, or composition patterns.

• Add a 50-200 character description to get_dag_status explaining what 'swarm state' includes (e.g., task status, agent assignments, execution timeline) and when to call it.
• Document the output schema for all three tools. Specify field names, types, and what each field represents. E.g., plan_swarm should return {task_id, status, steps[], estimated_duration}.
• Expand plan_swarm description to 100+ chars: 'Decompose a natural-language problem into a multi-agent execution plan. Returns task_id, step breakdown, and estimated completion time. Use this to understand how the swarm will solve the problem before execution.'
• Add error handling guidance to plan_swarm: 'If decomposition fails, returns {error, reason, suggestion}. Retryable if reason is 'timeout'; user-fixable if reason is 'ambiguous_problem'.'
• Add tool annotations: mark get_dag_status and validate_safety with readOnlyHint=true; mark plan_swarm with destructiveHint=true (since it executes actions).