[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-sandboxing":3},{"tag":4,"articles":10},{"id":5,"name":6,"slug":6,"article_count":7,"description_zh":8,"description_en":9},"88aaa510-19fc-4fd2-8289-fc399eaed3b1","sandboxing",3,"Sandboxing 指的是把代理人、程式碼或模型執行放進隔離環境，限制檔案、網路與系統權限。它常見於長流程 AI agent、企業自動化與雲端執行沙箱，重點在降低風險、提升可控性與部署彈性。","Sandboxing isolates code, agents, or model-driven workflows from the host system by restricting file, network, and OS access. It matters for enterprise AI agents, cloud execution, and generated code runs where tighter control reduces risk and improves deployment flexibility.",[11],{"id":12,"slug":13,"title":14,"summary":15,"category":16,"image_url":17,"cover_image":17,"language":18,"created_at":19},"b3e4d4ed-a225-456c-8550-8d26c26ce5e5","openai-agents-sdk-safer-enterprise-controls-zh","OpenAI Agents SDK 加強企業控管","OpenAI 為 Agents SDK 加上 sandboxing 與 harness 支援，讓企業能把長流程代理人放進更可控的環境，API 也維持標準計價。","ai-agent","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1776513829857-0otz.png","zh","2026-04-18T12:03:31.394038+00:00"]