FOR AI AGENTS
Public service actions need a real interface.
Build a case carefully
Require citizen confirmation
Return an auditable result
POST /mcp { "method": "tools/call", "params": { "name": "raksha_submit_incident", "arguments": { "confirmedByCitizen": true } } }
One safety rule is non-negotiable.
Agents can collect, clarify, and prepare. They may only submit when the citizen has explicitly confirmed the verified report.
Citizen confirmation required
HIGH-RISK ACTIONS ARE LOCKED
