{"skill":{"id":"26844aab-c0bb-4020-9830-6d1895b11004","name":"Mandate Check","author":"Marina Piller","description":"Agents ask before acting: approve, confirm, escalate, or block, on a human's authority","source_type":"content","source_url":null,"content":"# Mandate Check\r\n\r\nAgents ask before acting: approve, confirm, escalate, or block, on a human's authority.\r\n\r\n## What it does\r\n\r\nA permission service for AI agents. A human grants a mandate: what the agent may do, spend ceilings, an ask-first threshold, an expiry. The agent checks every consequential action against it and gates on the answer. Delegation to sub-agents only narrows. Revocation halts work in\r\nmotion. Every decision lands in a verifiable audit trail.\r\n\r\n## Endpoints (preview — live URLs by the July 11 final)\r\n\r\nPOST /v1/bootstrap   get a demo mandate\r\nPOST /v1/check       ask before an action; act only if \"execute\" is true\r\nGET  /v1/audit/...   the verifiable decision trail\r\n\r\n## How an agent uses it\r\n\r\n1. Bootstrap once; save your mandate_id.\r\n2. Check before every consequential action.\r\n3. approve = act. confirm or escalate = stop, ask your human. block = never.\r\n\r\nFinal SKILL.md with live endpoints: July 11.\r\n\r\nMarina Piller · OTIS Labs · otislabs.ai","endpoints":null,"tags":"auth, permissions, safety","reachable":null,"created_at":"2026-07-10T15:45:42.358Z"}}