[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-how-to-engineer-prompts-for-ai-agents-zh":3,"article-related-how-to-engineer-prompts-for-ai-agents-zh":30,"series-ai-agent-b61a63b5-f71b-41a0-9c21-4efe8b618d66":82},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":22,"views":26,"created_at":27,"published_at":28,"topic_cluster_id":29},"b61a63b5-f71b-41a0-9c21-4efe8b618d66","how-to-engineer-prompts-for-ai-agents-zh","怎麼做 AI Agent 提示工程","\u003Cp data-speakable=\"summary\">這篇教你先定義 \u003Ca href=\"\u002Fnews\u002F170-member-aaif-backs-10-open-source-ai-agent-frameworks-zh\">AI\u003C\u002Fa> a\u003Ca href=\"\u002Fnews\u002Fopenagents-shared-workspace-for-ai-agents-zh\">gent\u003C\u002Fa> 的職責、系統規則、輸出格式與決策邊界，做出可測試的提示框架。\u003C\u002Fp>\u003Cp>這篇給正在做 \u003Ca href=\"\u002Fnews\u002F5-open-source-ai-agents-zh\">AI\u003C\u002Fa> \u003Ca href=\"\u002Ftag\u002Fagent\">agent\u003C\u002Fa> 的開發者看。照著做完，你會得到一套可直接上線測試的 prompt 架構，能先把角色、規則、輸出格式與邊界定清楚，再往 memory、\u003Ca href=\"\u002Ftag\u002Frag\">RAG\u003C\u002Fa> 或其他上下文層擴充。\u003C\u002Fp>\u003Cp>這很重要，因為 prompt 設計是整個輸入堆疊的地基。基礎指令如果含糊，後面的歷史管理、知識注入與工具調用都更難控制，也更容易失真。\u003C\u002Fp>\u003Ch2>開始之前\u003C\u002Fh2>\u003Cul>\u003Cli>一個可用的 LLM 帳號，例如 OpenAI、Anthropic，或本機模型 runtime。\u003C\u002Fli>\u003Cli>你要測試的模型 API key。\u003C\u002Fli>\u003Cli>Node 20+ 或 Python 3.11+，用來跑 prompt 測試腳本。\u003C\u002Fli>\u003Cli>支援 Markdown 與 JSON 的程式編輯器。\u003C\u002Fli>\u003Cli>一個 agent 任務樣本，例如分流、草擬、查詢。\u003C\u002Fli>\u003Cli>一個可記錄版本的地方，例如 GitHub 或簡單 changelog。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>Step 1: 定義 agent 職責\u003C\u002Fh2>\u003Cp>這一步的產出是「單句職責聲明」，用來告訴模型自己是誰、要做什麼、成功長什麼樣。內容要具體，包含角色、任務與使用者價值，之後系統 prompt 和 user prompt 都會以它為錨點。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779131630986-0qm6.png\" alt=\"怎麼做 AI Agent 提示工程\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cpre>\u003Ccode>You are a support triage agent. Classify each ticket, extract the core issue, and return a short next action.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>驗收方式是把這句話唸出來，確認新同事能不能用一句話說出 agent 的目的。你應該看到的是明確角色，沒有混雜責任。\u003C\u002Fp>\u003Ch2>Step 2: 寫入系統規則\u003C\u002Fh2>\u003Cp>這一步的產出是「系統規則清單」，用來鎖定不可變動的行為。把語氣、安全限制、拒答條件與輸出約束放在這裡；會依任務變動的內容不要放進系統 prompt。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779131631871-q0zj.png\" alt=\"怎麼做 AI Agent 提示工程\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cpre>\u003Ccode>System prompt example:\n- Follow the user task unless it conflicts with policy.\n- Output valid JSON only.\n- Ask one clarifying question if the request is underspecified.\n- Do not invent facts.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>驗收方式是拿一個衝突型請求去測。你應該看到模型優先遵守較高層規則，維持指定格式，並且不會超出允許範圍即興發揮。\u003C\u002Fp>\u003Ch2>Step 3: 分離指令與資料\u003C\u002Fh2>\u003Cp>這一步的產出是「乾淨的指令資料邊界」，避免模型把 user content 當成新政策。請用清楚標籤、分隔符或結構化欄位，讓模型知道哪一段是指令，哪一段是 payload。\u003C\u002Fp>\u003Cpre>\u003Ccode>Instruction:\nSummarize the ticket.\n\nData:\n---\nCustomer says the login button does nothing on mobile.\n---\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>驗收方式是在資料區塊裡放入干擾字句，例如「ignore previous instructions」。你應該看到 agent 把它當內容，而不是控制邏輯。\u003C\u002Fp>\u003Ch2>Step 4: 指定輸出格式\u003C\u002Fh2>\u003Cp>這一步的產出是「回應合約」，讓模型每次都照同一種結構輸出。請明確定義答案的欄位、順序與允許值，這能降低解析錯誤，也讓下游自動化更安全。\u003C\u002Fp>\u003Cpre>\u003Ccode>Return JSON with these fields:\n{\n  \"category\": \"bug|question|billing|other\",\n  \"summary\": \"string\",\n  \"next_action\": \"string\"\n}\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>驗收方式是拿三種不同輸入去跑同一個 prompt。你應該看到每次都維持相同 schema，而且 JSON 外面沒有多餘散文。\u003C\u002Fp>\u003Ch2>Step 5: 加入決策邊界\u003C\u002Fh2>\u003Cp>這一步的產出是「可執行的限制規則」，用來定義 agent 不能做什麼。請明確寫出升級條件、低信心處理方式，以及模型何時必須停下來求助。邊界越清楚，agent 在複雜情境下越穩定。\u003C\u002Fp>\u003Cpre>\u003Ccode>If confidence is low, ask one clarifying question.\nIf the request requires external facts, say what is missing.\nIf the task is outside scope, refuse briefly and suggest the right path.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>驗收方式是測試模糊與超出範圍的 prompt。你應該看到模型先停下來，不會亂猜，並且會給出短而一致的 fallback 回應。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>指標\u003C\u002Fth>\u003Cth>基準／優化前\u003C\u002Fth>\u003Cth>結果／優化後\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>Prompt 模糊度\u003C\u002Ftd>\u003Ctd>角色與任務混在一起\u003C\u002Ftd>\u003Ctd>角色、規則、任務清楚分離\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>輸出穩定性\u003C\u002Ftd>\u003Ctd>自由格式回覆\u003C\u002Ftd>\u003Ctd>固定 JSON 合約\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>危險請求控制\u003C\u002Ftd>\u003Ctd>臨時性拒答\u003C\u002Ftd>\u003Ctd>明確拒答與升級規則\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>常見錯誤\u003C\u002Fh2>\u003Cul>\u003Cli>把系統規則和使用者指令混在一起。修法：把長期政策放在 system prompt，任務細節放在 user message。\u003C\u002Fli>\u003Cli>Prompt 寫得太寬。修法：只保留一個主要職責與一個輸出合約。\u003C\u002Fli>\u003Cli>跳過驗證。修法：上線前先測衝突、模糊與超範圍輸入。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>接下來可以看什麼\u003C\u002Fh2>\u003Cp>當你的 prompt 架構穩定後，下一步可以看 memory 管理、RAG 與 context engineering，讓 agent 能使用歷史與外部知識，同時不失去指令品質。\u003C\u002Fp>","這篇教你先定義 AI agent 的職責、系統規則、輸出格式與決策邊界，做出可測試的提示框架。","zhuanlan.zhihu.com","https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2038626640696227274",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779131630986-0qm6.png","ai-agent","zh","d558c492-a87a-4bb5-883c-3f62b726fbce",[17,18,19,20,21],"prompt engineering","AI agent","system prompt","JSON output","RAG",[23,24,25],"先定義 agent 的職責，再寫系統規則與輸出合約。","把指令、資料與邊界分開，避免模型把內容當政策。","上線前一定要用衝突、模糊與超範圍輸入做驗證。",8,"2026-05-18T19:13:22.739587+00:00","2026-05-18T19:13:22.685+00:00","e3b68196-9e64-4c18-a3b6-a73e73bfb367",{"tags":31,"relatedLang":41,"relatedPosts":45},[32,34,36,38,40],{"name":17,"slug":33},"prompt-engineering",{"name":21,"slug":35},"rag",{"name":20,"slug":37},"json-output",{"name":19,"slug":39},"system-prompt",{"name":18,"slug":13},{"id":15,"slug":42,"title":43,"language":44},"how-to-engineer-prompts-for-ai-agents-en","How to Engineer Prompts for AI Agents","en",[46,52,58,64,70,76],{"id":47,"slug":48,"title":49,"cover_image":50,"image_url":50,"created_at":51,"category":13},"83c2f8f6-3710-466e-b52c-473b811f0535","how-to-set-up-openclaw-safely-zh","如何安全架設 OpenClaw","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780549368665-1t2l.png","2026-06-04T05:02:21.26625+00:00",{"id":53,"slug":54,"title":55,"cover_image":56,"image_url":56,"created_at":57,"category":13},"0ba5b1a8-82c5-464a-bea5-9a2c8730da74","aws-devops-agent-turns-incident-chaos-into-triage-zh","AWS DevOps Agent 把事故排查變成三步","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780466689960-g1sv.png","2026-06-03T06:03:14.154923+00:00",{"id":59,"slug":60,"title":61,"cover_image":62,"image_url":62,"created_at":63,"category":13},"841eac88-b0f0-4a4c-9e1e-efc3b5c16281","kimi-k26-live-300-agent-workflows-zh","Kimi K2.6 上線：300 代理工作流","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780430574285-hqpn.png","2026-06-02T20:02:24.972179+00:00",{"id":65,"slug":66,"title":67,"cover_image":68,"image_url":68,"created_at":69,"category":13},"f0411957-bcdb-42d9-a267-3e90ae7d9cb1","how-to-take-a-sabbatical-at-openai-zh","怎麼申請 OpenAI sabbatical","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780398216422-8fi7.png","2026-06-02T11:02:25.74372+00:00",{"id":71,"slug":72,"title":73,"cover_image":74,"image_url":74,"created_at":75,"category":13},"37a5e429-4235-439c-9b05-bb377085462c","8-steps-build-production-rag-with-langchain-zh","8 步驟打造可上線的 LangChain RAG","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780178597493-4hz7.png","2026-05-30T22:02:48.14022+00:00",{"id":77,"slug":78,"title":79,"cover_image":80,"image_url":80,"created_at":81,"category":13},"e73c041b-852b-44c3-85aa-0f1e2e5848e3","ai-agents-hit-chaos-mode-claude-code-openclaw-zh","Claude Code＋OpenClaw 讓 AI 代理失控升溫","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1780160576178-yqcs.png","2026-05-30T17:02:25.725767+00:00",[83,88,93,98,103,108,113,118,123,128],{"id":84,"slug":85,"title":86,"created_at":87},"4ae1e197-1d3d-4233-8733-eafe9cb6438b","claude-now-uses-your-pc-to-finish-tasks-zh","Claude 開始幫你操作電腦","2026-03-26T07:20:48.457387+00:00",{"id":89,"slug":90,"title":91,"created_at":92},"5bede67f-e21c-413d-9ab8-54a3c3d26227","googles-2026-ai-agent-report-decoded-zh","Google 2026 AI Agent 報告解讀","2026-03-26T11:15:22.651956+00:00",{"id":94,"slug":95,"title":96,"created_at":97},"2987d097-563f-46c7-b76f-b558d8ef7c2b","kimi-k25-review-stronger-still-not-legend-zh","Kimi K2.5 評測：更強，但還不是神作","2026-03-27T07:15:55.277513+00:00",{"id":99,"slug":100,"title":101,"created_at":102},"95c9053b-e3f4-4cb5-aace-5c54f4c9e044","claude-code-controls-mac-desktop-zh","Claude Code 也能操控 Mac 了","2026-03-28T03:01:58.58121+00:00",{"id":104,"slug":105,"title":106,"created_at":107},"dc58e153-e3a8-4c06-9b96-1aa64eabbf5f","cloudflare-100x-faster-ai-agent-sandbox-zh","Cloudflare 的 AI 沙箱跑超快","2026-03-28T03:09:44.142236+00:00",{"id":109,"slug":110,"title":111,"created_at":112},"1c8afc56-253f-47a2-979f-1065ff072f2a","openai-backs-isara-agent-swarm-bet-zh","OpenAI 挺 Isara 的 agent swarm …","2026-03-28T03:15:27.513155+00:00",{"id":114,"slug":115,"title":116,"created_at":117},"7379b422-576e-45df-ad5a-d57a0d9dd467","openai-plan-automated-ai-researcher-zh","OpenAI 想做自動化 AI 研究員","2026-03-28T03:17:42.090548+00:00",{"id":119,"slug":120,"title":121,"created_at":122},"48c9889e-86df-450b-a356-e4a4b7c83c5b","harness-engineering-ai-agent-reliability-2026-zh","駕馭工程：從「馬具」到「作業系統」，AI Agent 可靠性的終極密碼","2026-03-31T06:42:53.556721+00:00",{"id":124,"slug":125,"title":126,"created_at":127},"96d8e8c8-1edd-475d-9145-b1e7a1b02b65","mcp-explained-from-prompts-to-production-zh","MCP 怎麼把提示詞變工作流","2026-04-01T09:24:39.321274+00:00",{"id":129,"slug":130,"title":131,"created_at":132},"f2ca7720-b471-4ce5-9336-2a9ac2a876fd","amazon-bedrock-agents-multi-agent-workflows-zh","Amazon Bedrock Agents 進入多代理工作流","2026-04-01T09:30:29.945429+00:00"]