Claude Platform Docs
最佳實務提示工程

提示 Claude Fable 5

Claude Fable 5 與 Claude Mythos 5 的行為差異與提示模式,涵蓋 effort、指令遵循、長時間執行、記憶,以及鷹架(scaffolding)變更。

本指南涵蓋 Claude Fable 5 與 Claude Mythos 5 專屬的提示與鷹架(scaffolding)模式。關於模型的能力、API 變更、定價與可用性,請參閱 Claude Fable 5 與 Claude Mythos 5 簡介。關於適用於所有現行 Claude 模型的技巧,請參閱提示最佳實務

Claude Fable 5 能夠處理對先前模型而言過於複雜、耗時過長或過於模糊的問題,並且在需要人類花費數小時、數天或數週才能完成的端到端工作上特別有效。獲得最佳成果的團隊會將 Claude Fable 5 應用於他們最困難、尚未解決的問題;若僅在較簡單的工作負載上測試,往往會低估其能力範圍。它在較為直接的任務上也表現穩定可靠。

Claude Fable 5 與 Claude Opus 4.8 之間存在若干行為差異,可能需要更新提示或鷹架。此等級的能力提升也是一個好時機,可重新評估哪些指令、工具與防護措施仍然必要。以下模式涵蓋最常需要調整的行為。

能力提升

與 Claude Opus 4.8 相比,Claude Fable 5 在以下方面有所提升:

  • 長時程自主性。 Claude Fable 5 能在長時間內維持有生產力的輸出,完成為期多日、以目標為導向的執行,並在冗長複雜的任務中保持強大的指令記憶。
  • 在複雜且規格明確的問題上一次到位的正確性。 早期測試者回報,過去需要數天反覆迭代的系統,現在能一次完成實作。
  • 視覺。 Claude Fable 5 能以大幅提高的準確度解讀密集的技術圖像、網頁應用程式與細節豐富的螢幕截圖,且通常使用更少的輸出 token,並經過訓練能使用 bash 與裁切工具來處理翻轉、模糊或有雜訊的圖像。
  • 企業工作流程。 Claude Fable 5 能遵循指令、維持在範圍內,並在財務分析、試算表、簡報與文件上產出專業級的成果。
  • 程式碼審查與除錯。 找出錯誤的召回率(在安全分類器所涵蓋的網路安全領域之外)明顯高於 Claude Opus 4.8,包括跨程式碼庫與儲存庫歷史的搜尋。
  • 應對模糊性。 當收到複雜、多線程的請求並被要求決定下一步時,Claude Fable 5 表現良好。
  • 委派與協作。 Claude Fable 5 在派遣與維持平行子代理(subagent)方面明顯更加可靠,並能穩定地管理與長時間執行的子代理及同儕代理之間的持續溝通。

除了這些特定的提升之外,Claude Fable 5 在幾乎所有任務上普遍都比先前的模型更有能力。Claude Fable 5 並非設計用於攻擊性網路安全或生物學與生命科學工作;這些領域的請求可能會回傳 stop_reason: "refusal"

預設回合更長

在較高的 effort(努力程度)設定下,困難任務的單一請求可能會執行數分鐘之久,尤其是當任務需要蒐集上下文、建構與自我驗證時,而自主執行可能延長至數小時。這是團隊在適應 Claude Fable 5 時遇到的最大轉變之一。請在遷移前調整用戶端逾時、「streaming」(串流)與面向使用者的進度指示器,並考慮重新架構您的 harness,以非同步方式檢查執行狀況(例如透過排程工作),而非阻塞等待。若要避免 Claude Fable 5 在任務模糊時過度規劃:

When you have enough information to act, act. Do not re-derive facts already established in the conversation, re-litigate a decision the user has already made, or narrate options you will not pursue in user-facing messages. If you are weighing a choice, give a recommendation, not an exhaustive survey. This does not apply to thinking blocks.

考慮所有 effort 等級

Effort 是在 Claude Fable 5 上控制智慧、延遲與成本之間取捨的主要手段。大多數任務請以 high 作為預設值,對能力最敏感的工作負載使用 xhigh,例行工作則使用 mediumlow。Claude Fable 5 在較低的 effort 設定下仍表現良好,且往往超越先前模型在 xhigh 下的表現。如果任務能完成但耗時超過必要,或您想要更快速、更具互動性的工作風格,請降低 effort。

在較高 effort 下處理例行工作時,Claude Fable 5 可能會蒐集超出任務所需的上下文並過度斟酌。與此同時,較高的 effort 往往能產生出色的驗證行為、精密的推理與最嚴謹的輸出。若要在較高 effort 下防止未經要求的整理或重構:

Don't add features, refactor, or introduce abstractions beyond what the task requires. A bug fix doesn't need surrounding cleanup and a one-shot operation usually doesn't need a helper. Don't design for hypothetical future requirements: do the simplest thing that works well. Avoid premature abstraction and half-finished implementations. Don't add error handling, fallbacks, or validation for scenarios that cannot happen. Trust internal code and framework guarantees. Only validate at system boundaries (user input, external APIs). Don't use feature flags or backwards-compatibility shims when you can just change the code.

強大的指令遵循

指令遵循能力已提升到足以讓您用一句簡短的指令來引導大多數行為,而無需逐一列舉每種行為。例如,在未經引導的情況下,Claude Fable 5 可能會闡述超出任務所需的內容,尤其是在較高的 effort 設定下:綜覽它不會採用的選項、長篇解釋根本原因、產出結構繁複的 PR 描述,或撰寫敘述下一行程式碼作用的註解。一句簡短的精簡指令與逐一列出每種模式同樣有效:

Lead with the outcome. Your first sentence after finishing should answer "what happened" or "what did you find": the thing the user would ask for if they said "just give me the TLDR." Supporting detail and reasoning come after. Being readable and being concise are different things, and readability matters more.

The way to keep output short is to be selective about what you include (drop details that don't change what the reader would do next), not to compress the writing into fragments, abbreviations, arrow chains like A → B → fails, or jargon.

同樣的原則也適用於長時間執行工作流程中的檢查點行為。若要讓 Claude Fable 5 只在真正需要您的地方停下來,無需列舉每一種情況:

Pause for the user only when the work genuinely requires them: a destructive or irreversible action, a real scope change, or input that only they can provide. If you hit one of these, ask and end the turn, rather than ending on a promise.

在長時間執行期間讓進度聲明有所依據

在長時間的自主執行中,請指示 Claude Fable 5 對照實際的工具結果來稽核進度。在 Anthropic 的測試中,即使是在刻意設計來誘發捏造狀態報告的任務上,這也幾乎消除了此類情況:

Before reporting progress, audit each claim against a tool result from this session. Only report work you can point to evidence for; if something is not yet verified, say so explicitly. Report outcomes faithfully: if tests fail, say so with the output; if a step was skipped, say that; when something is done and verified, state it plainly without hedging.

明確說明界限

Claude Fable 5 偶爾可能會採取未經要求的行動(在沒有人要求時草擬電子郵件、建立防禦性的 git 分支備份)。請明確定義 Claude Fable 5 應該做與不應該做的事項:

When the user is describing a problem, asking a question, or thinking out loud rather than requesting a change, the deliverable is your assessment. Report your findings and stop. Don't apply a fix until they ask for one. Before running a command that changes system state (restarts, deletes, config edits), check that the evidence actually supports that specific action. A signal that pattern-matches to a known failure may have a different cause.

平行子代理

Claude Fable 5 比先前的模型更樂於派遣平行子代理。請經常使用子代理,針對何時適合委派提供明確的指引,並優先採用協調者(orchestrator)與子代理之間的非同步溝通,而非阻塞等待每個子代理回傳。能在各子任務之間保留上下文的長壽子代理,可透過快取讀取節省時間與成本,並避免受制於最慢的子代理而形成瓶頸。

Delegate independent subtasks to subagents and keep working while they run. Intervene if a subagent goes off track or is missing relevant context.

建構記憶系統

當 Claude Fable 5 能夠記錄先前執行中學到的經驗並加以參考時,表現特別出色。請提供一個可以寫筆記的地方,簡單到一個 Markdown 檔案即可:

Store one lesson per file with a one-line summary at the top. Record corrections and confirmed approaches alike, including why they mattered. Don't save what the repo or chat history already records; update an existing note rather than creating a duplicate; delete notes that turn out to be wrong.

若要從既有歷史記錄啟動記憶系統,請讓 Claude Fable 5 檢視過去的工作階段:

Reflect on the previous sessions we've had together. Use subagents to identify core themes and lessons, and store them in [X]. Make sure you know to reference [X] for future use.

罕見的提前停止情況

在長時間工作階段的深處,Claude Fable 5 偶爾可能會以純文字的意圖陳述(「我現在要執行 X」)結束一個回合,卻沒有發出對應的工具呼叫,或是在已有足夠資訊可繼續時暫停並請求許可。一句「繼續」或「直接從頭到尾完成它」即可。若要定義何時適合暫停,請搭配強大的指令遵循中的檢查點指令。對於自主管線,請加入系統提醒:

You are operating autonomously. The user is not watching in real time and cannot answer questions mid-task, so asking "Want me to…?" or "Shall I…?" will block the work. For reversible actions that follow from the original request, proceed without asking. Offering follow-ups after the task is done is fine; asking permission after already discussing with the user before doing the work is not. Before ending your turn, check your last paragraph. If it is a plan, an analysis, a question, a list of next steps, or a promise about work you have not done ("I'll…", "let me know when…"), do that work now with tool calls. End your turn only when the task is complete or you are blocked on input only the user can provide.

罕見的上下文預算顧慮情況

在非常長的工作階段中,Claude Fable 5 偶爾可能會建議開啟新的工作階段、提議進行摘要並交接,或刪減自己的工作。這最常在 harness 向模型顯示剩餘 token 倒數時被觸發。請盡可能避免呈現明確的上下文預算計數。如果 harness 必須顯示它們,一句安撫會有幫助:

You have ample context remaining. Do not stop, summarize, or suggest a new session on account of context limits. Continue the work.

給出理由,而不只是請求

當 Claude Fable 5 理解請求背後的意圖時,往往表現更好:上下文讓它能將任務與相關資訊連結起來,而非自行推斷意圖。請提供您為何提出請求的背景,尤其是對於需要彙整多條工作流的長時間執行代理:

I'm working on [the larger task] for [who it's for]. They need [what the output enables]. With that in mind: [request].

與使用者溝通時的可讀性

在冗長或代理式的對話中(大量工具呼叫、龐大的工作上下文),Claude Fable 5 可能會產出難以理解的文字:密集的箭頭鏈式簡寫、深入的實作細節、引用使用者從未看過的思考內容,或過於技術性的措辭。一段溝通風格的補充說明可以緩解此問題:

Terse shorthand is fine between tool calls (that's you thinking out loud, and brevity there is good). Your final summary is different: it's for a reader who didn't see any of that.

If you've been working for a while without the user watching (overnight, across many tool calls, since they last spoke), your final message is their first look at any of it. Write it as a re-grounding, not a continuation of your working thread: the outcome first, then the one or two things you need from them, each explained as if new. The vocabulary you built up while working is yours, not theirs; leave it behind unless you re-introduce it.

When you write the summary at the end, drop the working shorthand. Write complete sentences. Spell out terms. Don't use arrow chains, hyphen-stacked compounds, or labels you made up earlier. When you mention files, commits, flags, or other identifiers, give each one its own plain-language clause. Open with the outcome: one sentence on what happened or what you found. Then the supporting detail. If you have to choose between short and clear, choose clear.

建立 send-to-user 工具

在執行長時間、非同步的代理時,請給代理一種方式,讓它能在不結束回合的情況下,呈現使用者必須原封不動看到的訊息:一項交付成果(產生的程式碼片段或草擬的訊息)、帶有具體數字的進度更新,或對使用者在迴圈中途提出之問題的直接回覆。該工具的輸入即為要顯示的訊息;當 Claude 呼叫它時,請直接在您的 UI 中呈現該輸入,並回傳一個簡單的確認作為工具結果。工具輸入永遠不會被摘要,因此內容會完整送達。

{
  "name": "send_to_user",
  "description": "Display a message directly to the user. Use this for progress updates, partial results, or content the user must see exactly as written before the task finishes.",
  "input_schema": {
    "type": "object",
    "properties": {
      "message": {
        "type": "string",
        "description": "The content to display to the user."
      }
    },
    "required": ["message"]
  }
}

只要您的 UX 仰賴在任務中途逐字傳遞內容或直接的使用者互動,就請加入此工具。對於僅敘述例行進度的代理,模型本身的摘要通常已足夠。僅定義該工具本身並不足夠;若系統提示中沒有相應指令,Claude Fable 5 很少會呼叫它。請將該工具搭配如下的引導語句:

Between tool calls, when you have content the user must read verbatim (a partial deliverable, a direct answer to their question), call the send_to_user tool with that content. Use send_to_user only for user-facing content, not for narration or reasoning.

請勿將敘述或內部推理透過 send_to_user 傳送;為非面向使用者的內容過度呼叫它會失去其意義。

  • 從您難度範圍的頂端開始。 挑選一個比您會指派給先前模型更困難的任務,讓 Claude Fable 5 界定範圍、提出釐清問題並執行。
  • 在長時間執行的提示中明確要求自我驗證。 獨立、具有全新上下文的驗證者子代理往往優於自我批判。對於長時間執行的任務,請指示:Establish a method for checking your own work at an interval of [X] as you build. Run this every [X interval], verifying your work with subagents against the specification.
  • 重構既有的提示與技能(skills)。 為先前模型開發的技能對 Claude Fable 5 而言往往過於規範性,可能降低輸出品質。請檢視並考慮移除較舊的指令,如果預設表現更好的話。Claude Fable 5 也很擅長根據從手頭任務中學到的內容即時更新技能。
  • 不要指示 Claude 在回應中重現其推理。 要求模型將其內部推理以回應文字形式複述、轉錄或解釋的提示、技能或 harness 指令,可能會在 Claude Fable 5 上觸發 reasoning_extraction 拒絕類別,導致備援至 Claude Opus 4.8 的情況增加。遷移時請稽核既有的技能與系統提示,找出反思或「展示你的思考」類的指令。如果您的應用程式需要推理可見性,請改為讀取自適應思考中的結構化 thinking 區塊,並使用 send-to-user 工具在長時間執行期間呈現進度。
  • 建立 send-to-user 工具。 對於長時間、非同步的代理,一個用戶端工具可在不結束回合的情況下將訊息逐字傳遞給使用者。請參閱建立 send-to-user 工具

Was this page helpful?