本指南涵蓋 Claude Fable 5 與 Claude Mythos 5 特有的提示與鷹架(scaffolding)模式。關於模型的能力、API 變更、定價與可用性,請參閱介紹 Claude Fable 5 與 Claude Mythos 5。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 能夠處理先前模型認為過於複雜、執行時間過長或過於模糊的問題,並且特別擅長需要人類花費數小時、數天或數週才能完成的端到端工作。獲得最佳成果的團隊會將 Claude Fable 5 應用於他們最困難的未解決問題;僅在較簡單的工作負載上測試它,往往會低估其能力範圍。它在較為直接的任務上也能可靠地執行。
Claude Fable 5 與 Claude Opus 4.8 相比有幾項行為差異,可能需要更新提示或鷹架。在這個層級的能力提升,也是重新評估哪些指令、工具與防護措施仍然必要的好時機。以下的模式涵蓋了最常需要調整的行為。
關於 Claude Fable 5 與 Claude Mythos 5 特有的 API 參數變更(僅支援自適應思考、僅輸出摘要式思考、無擴展思考預算、refusal 停止原因與後備處理),請參閱介紹 Claude Fable 5 與 Claude Mythos 5。
Claude Fable 5 會執行安全分類器,針對攻擊性網路安全技術(例如建構漏洞利用程式、惡意軟體或攻擊工具)、生物與生命科學內容(例如實驗室方法或分子機制),以及擷取模型的摘要式思考。良性的網路安全工作與有益的生命科學任務也可能觸發這些防護措施。若要自動重新路由被拒絕的請求,請設定伺服器端或用戶端後備至 Claude Opus 4.8。
與 Claude Opus 4.8 相比,Claude Fable 5 在以下方面有所提升:
除了這些具體的改進之外,Claude Fable 5 在幾乎所有任務上通常都比先前的模型更有能力。Claude Fable 5 並非為攻擊性網路安全或生物與生命科學工作而設計;這些領域的請求可能會回傳 stop_reason: "refusal"。
在較高的 effort 設定下,困難任務的單一請求可能會執行許多分鐘,特別是當任務需要收集上下文、建構與自我驗證時,而自主執行可能會延長至數小時。這是團隊在適應 Claude Fable 5 時遇到的最大轉變之一。在遷移之前,請調整用戶端逾時、串流與面向使用者的進度指示器,並考慮重新建構 harness 以非同步方式檢查執行狀態(例如透過排程工作),而非阻塞等待。為了避免 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.Effort 是 Claude Fable 5 上控制智慧、延遲與成本之間權衡的主要手段。大多數任務請使用 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 medium 或 low。Claude Fable 5 上較低的 effort 設定仍然表現良好,且通常超越先前模型的 xhigh 表現。如果任務能完成但花費的時間超過必要,或者您想要更快速、更具互動性的工作方式,請降低 effort。
在較高 effort 下處理例行工作時,Claude Fable 5 可能會收集超出任務所需的上下文並過度斟酌。同時,較高的 effort 通常會產生出色的驗證行為、精密的推理,以及最嚴謹的輸出。為了避免在較高 effort 下進行未被要求的整理或重構:
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.指令遵循能力已有足夠的提升,您可以用簡短的指令來引導大多數行為,而不需要逐一列舉每個行為。例如,在未受引導時,Claude Fable 5 可能會闡述超出任務所需的內容,特別是在較高的 effort 設定下:調查它不會採用的選項、冗長地解釋根本原因、產生結構過於繁複的 PR 描述,或撰寫描述下一行程式碼在做什麼的註解。一個簡短的簡潔性指令與列舉每個模式一樣有效:
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 只在真正需要您的時候停下來,不需要列舉每一種情況:
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.在長時間的自主執行中,請指示 Claude Fable 5 根據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虛構狀態報告的任務上,這幾乎完全消除了虛構的狀態報告:
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.Claude Fable 5 偶爾會採取未被要求的行動(在沒有人要求的情況下草擬電子郵件、建立防禦性的 git 分支備份)。請明確定義 Claude Fable 5 應該做什麼與不應該做什麼的限制:
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.Claude Fable 5 比先前的模型更容易派遣平行子代理。請經常使用子代理,提供關於何時適合委派的明確指引,並優先採用協調者與子代理之間的非同步通訊,而非阻塞等待每個子代理回傳。在子任務之間保留其上下文的長生命週期子代理,可透過快取讀取節省時間與成本,並避免受限於最慢的子代理。
Delegate independent subtasks to subagents and keep working while they run. Intervene if a subagent goes off track or is missing relevant context.當 Claude Fable 5 能夠記錄先前執行的經驗教訓並加以參考時,表現特別出色。請提供一個可以寫筆記的地方,簡單如一個 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.若要從現有歷史記錄引導建立記憶系統,請讓 Claude Fable 5 回顧過去的工作階段:
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.在長時間工作階段的深處,Claude Fable 5 偶爾會以純文字的意圖陳述(「我現在要執行 X」)結束回合,而沒有發出對應的工具呼叫,或者在已經有足夠資訊可以繼續時暫停並請求許可。一句「continue」或「go ahead and do it end to end」就足夠了。若要定義何時適合暫停,請將此與強大的指令遵循能力中的檢查點指令搭配使用。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.在非常長的工作階段中,Claude Fable 5 偶爾會建議開啟新的工作階段、提議進行摘要與交接,或縮減自己的工作。這最常在 harness 向模型顯示剩餘 token 倒數時被觸發。請盡可能避免顯示明確的上下文預算計數。如果 harness 必須顯示它們,一句安撫的話會有幫助:
You have ample context remaining. Do not stop, summarize, or suggest a new session on account of context limits. Continue the work.當 Claude Fable 5 理解請求背後的意圖時,往往表現得更好:上下文讓它能將任務與相關資訊連結起來,而不是自行推斷意圖。請提供您為何提出請求的背景資訊,特別是對於需要整合多個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].在延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 可能會產生難以理解的文字: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。加入溝通風格的補充說明可以緩解這個問題:
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.在執行長時間、非同步的代理時,請給代理一個方法,讓它能在不結束回合的情況下,向使用者呈現必須完全按照原文顯示的訊息:一個交付成果(產生的程式碼片段或草擬的訊息)、包含具體數字的進度更新,或對使用者在迴圈中途提出的問題的直接回覆。該工具的輸入就是要顯示的訊息;當 Claude 呼叫它時,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 依賴於在任務進行中逐字傳遞內容或直接的使用者互動,就加入這個工具。對於只敘述例行進度的代理,模型自己的摘要通常就足夠了。僅定義工具本身是不夠的;如果系統提示中沒有相關指令,Claude Fable 5 很少會呼叫它。請將該工具與引導語言搭配使用,例如:
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.不要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違背其目的。
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.reasoning_extraction 拒絕類別,導致更頻繁地後備至 Claude Opus 4.8。遷移時,請稽核現有的技能與系統提示中是否有反思或展示思考過程的指令。如果您的應用程式需要推理的可見性,請改為讀取自適應思考的結構化 thinking 區塊,並使用 send-to-user 工具在長時間執行期間呈現進度。Was this page helpful?