jevIMAGE
用 Jev 给 agent 加权限审批闸门
架构为 LLM 写、Jev 建议、应用校验权限,让被撤销授权的 agent 无法执行不可逆操作。
BuBBliK
@k1rallik29 天前收录原推发布于 2026.10.04

BuBBliK @k1rallik原推发布于 2026.10.04
this is insane. i genuinely don't understand why everyone building AI agents isn't shown this before giving their agent permission to change things they can't undo. imagine your agent says "deploy." every test passed. confidence is high. your app still blocks it. approval was revoked 30 seconds ago. LLM writes → Jev recommends → your app checks permission. that's the architecture this guide walks through:
这条是引用着另一条写的:x.com/mellon0x/status/2106107298418512123
x.com/k1rallik/status/2106434148433080641采集于 2026.10.04 10:04