verify_suggestion

Agents occasionally hallucinate tool names. verify_suggestion validates each suggestion against the graph and flags ones that don't exist or don't fit the declared use case.

Input

query················································································stringrequired

The user's original use case.

agent_suggestions················································································string[]required

Tool names the agent came up with on its own.

Example

{
  "query": "background jobs for a Node.js app",
  "agent_suggestions": ["BullMQ", "ImaginaryQueue", "Agenda"]
}

{
  "verdicts": [
    { "tool": "BullMQ",         "exists": true,  "fits": true,  "confidence": 0.92 },
    { "tool": "ImaginaryQueue", "exists": false, "fits": false, "confidence": 0.00,
      "note": "No matching tool in the ToolCairn graph." },
    { "tool": "Agenda",         "exists": true,  "fits": true,  "confidence": 0.78,
      "note": "Maintenance-tier; active alternatives are BullMQ and pg-boss." }
  ]
}

Usage pattern

Run verify_suggestion before telling the user about tools the agent generated freely. If any verdict returns exists: false, fall back to search_tools.