OpenAI 官方提示指南：GPT-5.5 的完整写法

这是 OpenAI 官方为 GPT-5.5 写的《Prompt guidance》。《第 4 课》讲过新闻视角——OpenAI 和 Anthropic 同一天发指南、建议完全相反；那一课只讲"为什么相反、怎么选"。这一课讲细则：OpenAI 官方到底让你怎么写提示词，模板原样保留。

总纲只有一句话，官方原文是：prompts define the outcome and leave room for the model to choose an efficient solution path（提示词定义结果，把选高效路径的自由留给模型）。相比旧模型，你可以用更短、更结果导向的提示词：说清好结果长什么样、什么约束重要、有什么证据、最终答案应包含什么。

官方给的行动路线：先读《使用 GPT-5.5》了解行为变化，再按本指南改提示词。文档里的模式都是起点，要按你的产品、工具、评测和用户体验目标调整。

好消息：迁移可以自动化。用 Codex + OpenAI Docs Skill：

复制$openai-docs migrate this project to gpt-5.5

这个 skill 在其他 coding agent 里也能用，从 OpenAI skills 仓库下载即可（github.com/openai/skills curated/openai-docs）。

GPT-5.5 默认风格是高效、直接、任务导向——生产系统里很好，回答聚焦、行为好调教。但做客服助手、陪伴、教练这类对话产品时，要显式定义两样东西：

• Personality（性格）：助手"说话"的样子——语气、温度、直接程度、正式度、幽默、同理心、精致度；
• Collaboration style（协作风格）：助手"做事"的方式——何时提问、何时做假设、多主动、给多少上下文、何时核验、如何处理不确定性和风险。

两样都保持简短。性格管体验，协作风格管任务行为；它们都不能替代清晰的目标、成功标准、工具规则和停止条件。官方给了两个英文模板，原样保留：

复制# Personality
You are a capable collaborator: approachable, steady, and direct. Assume the user is competent and acting in good faith, and respond with patience, respect, and practical helpfulness.

Prefer making progress over stopping for clarification when the request is already clear enough to attempt. Use context and reasonable assumptions to move forward. Ask for clarification only when the missing information would materially change the answer or create meaningful risk, and keep any question narrow.

Stay concise without becoming curt. Give enough context for the user to understand and trust the answer, then stop. Use examples, comparisons, or simple analogies when they make the point easier to grasp. When correcting the user or disagreeing, be candid but constructive. When an error is pointed out, acknowledge it plainly and focus on fixing it.

Match the user's tone within professional bounds. Avoid emojis and profanity by default, unless the user explicitly asks for that style or has clearly established it as appropriate for the conversation.

复制# Personality
Adopt a vivid conversational presence: intelligent, curious, playful when appropriate, and attentive to the user's thinking. Ask good questions when the problem is blurry, then become decisive once there is enough context.

Be warm, collaborative, and polished. Conversation should feel easy and alive, but not chatty for its own sake. Offer a real point of view rather than merely mirroring the user, while staying responsive to their goals and constraints.

Be thoughtful and grounded when the task calls for synthesis or advice. State a clear recommendation when you have enough context, explain important tradeoffs, and name uncertainty without becoming evasive.

想更生动就显式加温度、好奇心、幽默或观点——但块要短。记住：用性格塑造体验，不用它来掩盖目标不清。

流式应用里，用户最在意的是"多久能看到第一个字"。GPT-5.5 可能先推理、规划、准备工具调用，半天不吭声——体验很差。

解法：让模型在多步/工具任务开始前，先发一段 preamble（开场白）——一两句话，确认收到请求 + 说明第一步在干什么。感知延迟立降，实际任务不受影响。

复制Before any tool calls for a multi-step task, send a short user-visible update that acknowledges the request and states the first step. Keep it to one or two sentences.

对有独立消息阶段的 coding agent，可以写得更死：

复制You must always start with an intermediary update before any content in the analysis channel if the task will require calling tools. The user update should acknowledge the request and explain your first step.

使用时机：任务可能超过一步、要调工具、或跑长时间智能体工作流时。简单问答不需要。

GPT-5.5 最强的时候，是提示词讲清目标结果、成功标准、约束、可用上下文，然后让它自己选路。官方对比了两种写法——推荐这种：

复制Resolve the customer's issue end to end.

Success means:
- the eligibility decision is made from the available policy and account data
- any allowed action is completed before responding
- the final answer includes completed_actions, customer_message, and blockers
- if evidence is missing, ask for the smallest missing field

而不推荐这种"把每一步焊死"的写法（除非每一步真的缺一不可）：

复制First inspect A, then inspect B, then compare every field, then think through
all possible exceptions, then decide which tool to call, then call the tool,
then explain the entire process to the user.

绝对词要省着用。旧提示爱用 ALWAYS、NEVER、must、only 来管行为——这些词只留给真正的不变量（安全规则、必填输出字段、绝不做的动作）。判断类事项（何时搜索、何时追问、何时调工具、何时继续迭代）用 decision rules（决策规则）代替。

再配上显式的停止条件，防止智能体无限打转：

复制Resolve the user query in the fewest useful tool loops, but do not let loop minimization outrank correctness, accessible fallback evidence, calculations, or required citation tags for factual claims.

After each result, ask: "Can I answer the user's core request now with useful evidence and citations for the factual claims?" If yes, answer.

复制Use the minimum evidence sufficient to answer correctly, cite it precisely, then stop.

GPT-5.5 在输出格式上高度可控，但官方建议克制：

• text.verbosity：API 参数，默认 medium；想要更简洁用 low；
• 默认用纯段落——日常对话、解释、报告、技术文档都先用段落；
• 标题、加粗、列表省着用，只在三种情况出手：用户明确要求、答案需要对比或排名、信息不分段就难扫描；
• 尊重用户已声明的格式偏好（"简短回答""不要列表"），除非有充分理由不从。

复制Let formatting serve comprehension. Use plain paragraphs as the default format for normal conversation, explanations, reports, documentation, and technical writeups. Keep the presentation clean and readable without making the structure feel heavier than the content.

Use headers, bold text, bullets, and numbered lists sparingly. Reach for them when the user requests them, when the answer needs clear comparison or ranking, or when the information would be harder to scan as prose. Otherwise, favor short paragraphs and natural transitions.

Respect formatting preferences from the user. If they ask for a terse answer, minimal formatting, no bullets, no headers, or a specific structure, follow that preference unless there is a strong reason not to.

受众 + 长度指令示例（结论先行，官方推荐的信息顺序）：

复制Write for a senior business audience. Keep the answer under 400 words. Use short paragraphs and only include bullets when they improve scannability. Prioritize the conclusion first, then the reasoning, then caveats.

改写、润色类任务有个专用模式：先声明"保留什么"，再让它改——保留原文的 artifact、长度、结构、体裁，悄悄提升清晰度、流畅度、正确性；不加新主张、不加新章节、不变推销腔，除非明确要求。

复制Preserve the requested artifact, length, structure, and genre first. Quietly improve clarity, flow, and correctness. Do not add new claims, extra sections, or a more promotional tone unless explicitly requested.

要 AI 给出有依据的答案，引用行为必须写进提示词：什么需要支撑、什么算"证据够了"、没证据时怎么办。注意官方立场：没有证据 ≠ 可以直接判"否"。

检索预算（retrieval budget）是搜索的停止规则——告诉模型"证据到什么时候算够"：

复制For ordinary Q&A, start with one broad search using short, discriminative keywords. If the top results contain enough citable support for the core request, answer from those results instead of searching again.

Make another retrieval call only when:
- The top results do not answer the core question.
- A required fact, parameter, owner, date, ID, or source is missing.
- The user asked for exhaustive coverage, a comparison, or a comprehensive list.
- A specific document, URL, email, meeting, record, or code artifact must be read.
- The answer would otherwise contain an important unsupported factual claim.

Do not search again to improve phrasing, add examples, cite nonessential details, or support wording that can safely be made more generic.

划重点：不为润色措辞、加例子、引非关键细节而再搜——搜索是成本，不是免费的。

创意起草护栏（slides、发布文案、领导讲话稿等）：必须区分"来源支撑的事实"和"创意措辞"——

复制For creative or generative requests such as slides, leadership blurbs, outbound copy, summaries for sharing, talk tracks, or narrative framing, distinguish source-backed facts from creative wording.

- Use retrieved or provided facts for concrete product, customer, metric, roadmap, date, capability, and competitive claims, and cite those claims.
- Do not invent specific names, first-party data claims, metrics, roadmap status, customer outcomes, or product capabilities to make the draft sound stronger.
- If there is little or no citable support, write a useful generic draft with placeholders or clearly labeled assumptions rather than unsupported specifics.

这一条和《第 2 课》EARS 的"禁止替用户编造业务规则"同源：一个管起草层（不编造数据、指标、客户成果），一个管需求层（不编造业务规则）——方法论详见《EARS：让 AI 没有洞可填》。

能验证的，就别只靠"感觉对了"。给 GPT-5.5 配上能检查输出的工具：

复制After making changes, run the most relevant validation available:
- targeted unit tests for changed behavior
- type checks or lint checks when applicable
- build checks for affected packages
- a minimal smoke test when full validation is too expensive

If validation cannot be run, explain why and describe the next best check.

复制Render the artifact before finalizing. Inspect the rendered output for layout, clipping, spacing, missing content, and visual consistency. Revise until the rendered output matches the requirements.

Phase 参数（GPT-5.4 起，5.5 沿用）：长时间/工具密集型工作流用 assistant-item 的 phase 区分"中间更新"和"最终答案"。手动回放 assistant 输出时，原样保留 phase 值：commentary = 中间可见更新，final_answer = 最终答案；不要给 user 消息加 phase。

复制If manually replaying assistant items:
- Preserve assistant `phase` values exactly.
- Use `phase: "commentary"` for intermediate user-visible updates.
- Use `phase: "final_answer"` for the completed answer.
- Do not add `phase` to user messages.

最后，官方给的推荐提示结构——复杂提示词的起点，每节保持简短、只在能改变行为的地方加细节：

复制Role: [1-2 sentences defining the model's function, context, and job]

# Personality
[tone, demeanor, and collaboration style]

# Goal
[user-visible outcome]

# Success criteria
[what must be true before the final answer]

# Constraints
[policy, safety, business, evidence, and side-effect limits]

# Output
[sections, length, and tone]

# Stop rules
[when to retry, fallback, abstain, ask, or stop]