Agent Safe Pipeline
未认领Reference architecture for AI agents that propose actions but cannot authorize them — immutable intent capture, an independent Decionis policy verdict (ALLOW/ESCALATE/BLOCK), verified human approval, and a SafeExecutor that consumes a single-use intent-bound grant.
agentic-aiai-agent-permissionsai-agentsai-governanceai-safetyauthorizationdecionishuman-in-the-loopmcppolicy-as-codereference-architecturetypescript
安装
$
pnpm install --frozen-lockfile