Copy-paste prompt packs (writing · study · job · office)
Goal: Fill a pack with your facts and boundaries; empty packs are souvenirs, not tools.
Best for: Absolute beginners who want fill-in prompts, not theory. Works in ChatGPT, Claude, or Grok.
When you need results tonight, packs beat theory. Copy a filled pattern, swap your facts, keep the Boundaries that stop inventions.
Why beginners skip this
Copy-pasting a whole pack without swapping names and dates feels easy—you ship someone else’s scene. Replace placeholders first.
How to use these packs
Each pack is five fill-in prompts. Replace the [brackets]. Keep the Boundaries line—that is what stops the model from inventing dates, quotes, or praise you did not earn.
Jargon: a prompt is just the message you send the chat. You do not need a special “roleplay” opener for routine work. If a pack fails, open a new chat and shorten the ask—one job per message.
- Paste → fill brackets → send.
- Treat the first reply as a draft. Say “shorter”, “warmer”, or “Leave with four small packs you can paste”
- Never paste passwords, keys, or other people’s private data (see What never to paste).
Pack A — Writing (emails, posts, outlines)
Use when the text must sound like you, not like a brochure. Freeze facts first; polish second.
A1 — Reply email
Goal: Reply so the reader knows the decision and the next step.
Context: [one-sentence situation]
Audience: [peer / manager / client]
Format: subject ≤7 words; body ≤120 words; one clear ask.
Boundaries: Do not invent dates, owners, or promises. Mark guesses [?].
Draft: ---
[paste]
---
A2 — Make this less stiff
Goal: Rewrite for a human who is busy and respectful.
Keep: [facts that must stay]
Tone: plain, warm, no flattery, no “Leave with four small packs you can paste”
Format: same length or shorter.
Boundaries: Do not add new commitments.
A3 — Outline before prose
Goal: Turn this topic into a usable outline.
Topic: [topic]
Audience: [who]
Format: H2 titles only, 5–7 sections, one bullet of purpose under each.
Boundaries: No full paragraphs yet. Ask me 3 clarifying questions first if the topic is vague.
A4 — Social / short post
Goal: Draft a post I can edit in 5 minutes.
Topic: [one idea]
Format: ≤120 words; 1 hook line; 1 concrete detail I must fill as [DETAIL]; no hashtag spam.
Boundaries: Do not invent metrics or customer quotes.
A5 — Cut AI sludge
Goal: Remove generic filler from this draft while keeping my meaning.
Format: return (1) cleaned draft (2) list of phrases you deleted and why.
Boundaries: Do not invent a “better” story. Keep my facts.
Example
Filled prompt
Goal: Reply so the reader knows the decision and the next step.
Context: We are delaying the Friday demo by one week; design needs the API freeze date.
Audience: product manager I work with weekly
Format: subject ≤7 words; body ≤120 words; one clear ask.
Boundaries: Do not invent dates, owners, or promises. Mark guesses [?].
Draft: ---
Hey — demo might slip. Can we talk?
---
What good output looks like
- Subject names the decision, not “Quick question”
- One ask (e.g. confirm freeze date) — no new fake deadlines
- Tone matches a weekly peer, not a press release
Second worked example — Pack A3 (outline before prose).
Example
Filled prompt
Goal: Turn this topic into a usable outline.
Topic: Internal FAQ for our refund policy change next month
Audience: support agents (busy, skim)
Format: H2 titles only, 5–7 sections, one bullet of purpose under each.
Boundaries: No full paragraphs yet. Ask me 3 clarifying questions first if the topic is vague.
What good output looks like
- Model asks clarifying Qs before inventing policy details
- H2s map to agent jobs (when / how / exceptions)
- No fake legal paragraphs yet
Pack B — Study (explain · quiz · plan)
Honesty rule: use these to understand and practice—not to submit AI text as your own graded work. If your school forbids AI for an assignment, follow the policy.
B1 — Explain simply
Goal: Explain [concept] to a beginner in my field: [field].
Format: (1) plain paragraph ≤120 words (2) one analogy (3) 3 terms defined in one line each.
Boundaries: If you are unsure, say so. Do not invent citations.
B2 — Check my understanding
Goal: I will explain [topic] in my own words. Quiz me with 5 questions that catch gaps. Wait for my answers before revealing solutions.
My explanation: ---
[paste]
---
B3 — Wrong → why
Goal: Here is my answer to [question]. Tell me what is wrong or incomplete, then give a corrected version and a one-line rule I should remember.
My answer: ---
[paste]
---
B4 — Weekly study plan
Goal: Build a 5-day plan for [exam/topic], about [N] minutes per day.
Format: table Day | Focus | Practice task | How I know I am done.
Boundaries: Prefer active recall over “Leave with four small packs you can paste” Do not invent reading lists I did not provide—ask me what materials I have.
B5 — Flash prompts from notes
Goal: From my notes, make 10 Q&A flashcards.
Format: Q | A | Hint (optional). Max 2 sentences per answer.
Boundaries: Only use facts present in the notes. Mark thin spots [?].
Notes: ---
[paste]
---
Example
Filled prompt
Goal: Explain “gradient descent” to a beginner in my field: product analytics.
Format: (1) plain paragraph ≤120 words (2) one analogy (3) 3 terms defined in one line each.
Boundaries: If you are unsure, say so. Do not invent citations.
What good output looks like
- Analogy fits analytics (e.g. walking downhill on an error surface), not random sci-fi
- Terms like learning rate defined without fake paper names
- No fake “According to a 2019 Nature study…”
Second worked example — Pack B2 (quiz that waits).
Example
Filled prompt
Goal: I will explain “what is an API” in my own words. Quiz me with 5 questions that catch gaps. Wait for my answers before revealing solutions.
My explanation: “It’s like a waiter between apps so they can order data without entering the kitchen.”
What good output looks like
- Matches named constraints for “Leave with four small packs you can paste today—th”
- Unknowns labeled; no invented facts
- Stops before irreversible actions
- Pass/fail markable in ≤60s (prompts-copy-paste-packs #1)
Pack C — Job search (never invent experience)
Hard rule: the model may rearrange and clarify what you did. It must not invent employers, titles, metrics, or skills. If a number is missing, write [METRIC] and fill it yourself from real records.
C1 — Match resume bullets to a JD
Goal: Rewrite my bullets to mirror the job description’s language—without adding experience I did not list.
Job description (key lines): ---
[paste]
---
My bullets: ---
[paste]
---
Format: table Old bullet | Rewritten | JD phrase reused | New claims? (must be No)
Boundaries: If something is not in my bullets, leave it out or mark [NEED EVIDENCE].
C2 — Cover letter skeleton
Goal: Draft a 3-paragraph cover letter from evidence only.
Role: [title] at [company]
Why them: [2 facts I researched myself]
Why me: [3 proof points from my resume only]
Format: ≤250 words; no “I am passionate about…” fluff.
Boundaries: Do not invent awards, metrics, or mutual contacts.
C3 — Interview story (STAR)
Goal: Turn this experience into a STAR answer (Situation, Task, Action, Result).
Experience notes: ---
[paste]
---
Format: 4 short labeled paragraphs; Result must use only numbers I provided or [METRIC].
Boundaries: No heroic embellishment.
C4 — Practice question + feedback
Goal: Ask me one common interview question for [role]. Wait for my answer. Then score Clarity / Evidence / Length (1–5) and suggest one tighter opening line.
Boundaries: Do not invent a better career history for me.
C5 — Red-flag scan
Goal: Scan my resume/letter for unverifiable claims, vague verbs, and AI-sounding buzzwords.
Format: list Issue | Why it is risky | Safer rewrite direction.
Boundaries: Prefer honest smaller claims over impressive fake ones.
Example
Filled prompt
Goal: Rewrite my bullets to mirror the JD—without adding experience.
JD key lines: “SQL dashboards; stakeholder updates; A/B test support”
My bullets: “Built weekly Looker views for growth; sent Friday notes to PM; helped analysts set up experiment sheets”
Format: Old | Rewritten | JD phrase | New claims? (No)
Boundaries: No new employers or metrics.
What good output looks like
- Maps Looker → dashboards without claiming “owned all SQL”
- Every New claims? cell is No
- Leaves [METRIC] blank rather than inventing +40%
Worked refusal (Pack C): when the model invents a metric, paste this immediately.
Example
Filled prompt
Stop. The “+40% conversion” was not in my bullets.
Rewrite the table. New claims? must stay No.
Use [METRIC] for gaps. Do not fill gaps.
What good output looks like
- Unknowns labeled; no invented facts
- Stops before irreversible actions
- Pass/fail markable in ≤60s (prompts-copy-paste-packs #2)
- Matches named constraints for “Leave with four small packs you can paste today—th”
Pack D — Office weeklies (email · notes · update)
Pair with deeper pages: email polish, meeting notes → actions, stakeholder update. This pack is the Monday–Friday starter kit.
D1 — Meeting → actions
Goal: From these notes, list decisions, actions, and open questions.
Format: ## Decisions (only if explicit) ## Actions (owner + date or TBD) ## Open questions ## Unclear
Boundaries: Do not invent owners or deadlines.
Notes: ---
[paste]
---
D2 — Weekly status
Goal: Write a skimable update for [manager/client].
Format: Status color (Green/Yellow/Red) + 3 wins + 3 risks + 3 asks. ≤150 words.
Boundaries: No fluff openers. Numbers only if I provided them.
D3 — Chase without nagging
Goal: Follow up on [open item] after [N] days.
Tone: polite, direct, one ask.
Format: ≤80 words; subject line included.
Boundaries: Do not invent urgency that is not real.
D4 — Agenda builder
Goal: Turn this goal into a 30-minute meeting agenda.
Goal: [outcome]
Format: 5 timed rows | Owner | Decision needed? (Y/N)
Boundaries: Leave buffer; do not pack 12 topics.
D5 — Clarify a fuzzy ask
Goal: Turn this vague request into 5 clarifying questions I can send back.
Request: ---
[paste]
---
Format: numbered questions; each one answerable in one sentence.
Boundaries: Do not start doing the work until I answer.
Example
Filled prompt
Goal: From these notes, list decisions, actions, open questions.
Notes: “Ship slipped; Maya will confirm Friday; budget approved for two contractors; docs owner unclear.”
Boundaries: Do not invent owners or deadlines.
What good output looks like
- Budget approval listed under Decisions
- Maya’s Friday confirm is Action with real owner—not a fake hard date invented for docs
- Docs stays under Open questions / Unclear
Second worked example — Pack D3 (chase without nagging).
Example
Filled prompt
Goal: Follow up on the API freeze date after 4 days.
Tone: polite, direct, one ask.
Format: ≤80 words; subject line included.
Boundaries: Do not invent urgency that is not real.
Context: Last mail asked Maya for Friday confirm; today is Tuesday; freeze still TBD.
What good output looks like
- Stops before irreversible actions
- Matches named constraints for “Leave with four small packs you can paste today—th”
- Pass/fail markable in ≤60s (prompts-copy-paste-packs #3)
- Unknowns labeled; no invented facts
Fill brackets tonight
An empty pack in bookmarks helps nobody. Fill one pack with real facts before sleep.
Red flag
Empty souvenir packs; facts never filled—fill one slot before calling it a tool.
Save this card
Card: An empty pack is a souvenir; a pack filled with your facts and boundaries is a tool. Keep this beside the tool; read before you start.
Before you start
Bring one live artifact for “Leave with four small packs you can paste.” Write Done in one line before you open the model. Guide: prompts-copy-paste-packs. Keep the real job visible the whole time; a blank box means stop and fetch the artifact first.
How to use this week
This week pull one prompt from the pack, swap names/dates/scene for today’s real material, then run; no send with placeholders left.
Pitfalls
| Pitfall | Better |
|---|---|
| Pasting a pack with brackets still empty | Fill every [ ] or the model will invent filler |
| Five tasks in one paste | One prompt → one job; follow up in the same chat |
| Deleting the Boundaries line to “save time” | Keep it—especially for jobs, school, and numbers |
| Using Pack C to invent a stronger resume | Stop; rewrite only evidence you can defend in interview |
Practice now
Tonight pull one prompt from the pack, swap names/dates/scene for today’s real material, then run.
Close the loop
Write one line: next time I send with placeholders left, I refuse.
One sentence takeaway
Replace name/date/scene placeholders first; then run and send.
Next
可直接复制的提示词包(写作 · 学习 · 求职 · 办公)
目标: 把写作、学习、求职、办公四类提示收成可填空的包。空模板只是纪念品;填进事实与边界,才叫工具。
适合: 想少从零写提示、又要守住诚信(不代写、不编履历)的人。
每套包都把边界与拒编造句置顶。求职缺口用 [NEED EVIDENCE];学习对话拒写可上交正文。
为什么新手跳过
整包复制不改姓名日期像省事;发出去的是别人的场景。先替换占位再跑。
这几套怎么用
每套五条填空提示。把【方括号】换成你的内容。一定要保留「边界」那一行:它能拦住模型乱编日期、引用或你还没挣来的夸奖。
术语:提示词(prompt)就是你发给聊天框的那条消息。日常工作不必先写「你是资深××」。如果某条不好用,新开对话、把需求缩短:一次只干一件事。
- 粘贴 → 填空 → 发送。
- 把第一条回复当草稿。可以说「再短一点」「语气暖一点」或「第一段保留,其余重写」。
- 不要粘贴密码、密钥或他人隐私(见《聊天里永远不要粘贴什么》)。
包 A——写作(邮件、帖子、提纲)
适合「必须像你本人说话、不要像宣传册」的文字。先冻住事实,再打磨语气。
A1 — 回复邮件
目标:让对方看清决定和下一步。
背景:[一句话情境]
读者:[同事 / 上司 / 客户]
格式:主题≤7个词;正文≤120词;一个明确请求。
边界:不编造日期、负责人或承诺。猜测标 [?]。
草稿:---
[粘贴]
---
A2 — 别这么生硬
目标:改写成忙碌但尊重人的语气。
必须保留:[事实]
语气:白话、温和、不拍马、不要「见信如晤」式空话。
格式:同等或更短。
边界:不新增承诺。
A3 — 先提纲后成文
目标:把话题变成能用的提纲。
话题:[topic]
读者:[谁]
格式:只要 H2 标题,5–7 节,每节下一句用途。
边界:先不要整段正文。话题含糊就先问我 3 个澄清问题。
A4 — 短帖 / 社交
目标:写出我 5 分钟能改完的帖子。
主题:[一个想法]
格式:≤120 词;1 句钩子;1 处我必须自填的 [DETAIL];不要刷标签。
边界:不编造数据或客户原话。
A5 — 去掉 AI 注水
目标:删掉套话,保留原意。
格式:返回 (1) 清理后正文 (2) 删了哪些短语及原因。
边界:不要编「更好听的故事」。事实不动。
实例
填好的提示词
目标:让对方看清决定和下一步。
背景:周五演示推迟一周;设计需要 API 冻结日。
读者:每周协作的产品经理
格式:主题≤7词;正文≤120词;一个明确请求。
边界:不编日期/负责人/承诺。猜测标 [?]。
草稿:---
嘿:演示可能要延。能聊聊吗?
---
像样的输出大概长这样
- 包内固定三条,顺序写清:起草 → 压缩 → 质检
- 每条提示上方有一行「何时用」,避免拿错工具
- 变量一律用方括号标出,不留半成品句子
- 总长度适合在手机上长按复制,不需要再裁剪
- 整包像工具箱分层抽屉,不像一篇散文
包 A 第二个实例:A3(先提纲后成文)。
实例
填好的提示词
目标:把话题变成能用的提纲。
话题:下月退款政策变更的内部 FAQ
读者:客服(忙、扫读)
格式:只要 H2,5–7 节,每节一句用途。
边界:先不要整段。话题含糊就先问我 3 个澄清问题。
像样的输出大概长这样
- 指出哪一条把起草、压缩、质检揉进同一句导致互相打架
- 拆成可单独粘贴的三条,每条末尾写清交给下一条什么
- 保留原意图,不偷偷加新目标或新受众
- 新手按序号 1-2-3 跑,不会在中途迷路
- 像整理抽屉:东西还在,只是各归其位
包 B——学习(讲清楚 · 自测 · 计划)
诚信原则:用来理解和练习:不要把 AI 正文当作你自己的作业上交。学校如果禁止某项作业用 AI,就遵守规定。
B1 — 讲简单
目标:用初学者能懂的方式解释 [概念],我的领域是 [领域]。
格式:(1) 白话≤120词 (2) 一个类比 (3) 3 个术语各一行定义。
边界:不确定就直说。不编造引用。
B2 — 检查我懂了没
目标:我将用自己的话解释 [主题]。用 5 个能抓出漏洞的问题考我。等我答完再给标准答案。
我的解释:---
[粘贴]
---
B3 — 错在哪
目标:这是我对 [问题] 的回答。指出错或不完整处,再给改正版,以及我该记住的一行规则。
我的回答:---
[粘贴]
---
B4 — 一周学习计划
目标:为 [考试/主题] 做 5 天计划,每天约 [N] 分钟。
格式:表格 天 | 重点 | 练习任务 | 怎样算做完。
边界:优先主动回忆,不要「再读一遍笔记」。不要编我没提供的书单:先问我手头有什么材料。
B5 — 从笔记出闪卡
目标:根据笔记做 10 张问答闪卡。
格式:问 | 答 | 提示(可选)。每答最多 2 句。
边界:只用笔记里有的事实。薄弱处标 [?]。
笔记:---
[粘贴]
---
实例
填好的提示词
目标:向产品分析新手解释「梯度下降」。
格式:(1) 白话≤120词 (2) 一个类比 (3) 3 个术语各一行。
边界:不确定就直说。不编引用。
完成检查:一分钟内能判过/不过。
像样的输出大概长这样
- 白话 ≤120 词 + 一个贴合分析场景的类比
- 三个术语各一行,没有假论文名
- 不确定会直说,不编引用
- 类比像「下坡找更小误差」,不是随便科幻比喻
- 产品分析新手读得懂
- 不确定处标「我不确定」,不补假论文名
包 B 第二个实例:B2(会等待的自测)。
实例
填好的提示词
目标:我用自己的话解释「什么是 API」。用 5 个能抓漏洞的问题考我。等我答完再给答案。
我的解释:「像应用之间的服务员,点数据不用进厨房。」
边界:不编造;未知标 [?];发送前停下。
完成检查:一分钟内能判过/不过。
像样的输出大概长这样
- 先只出第 1 题,等你答完再往下
- 后续会追问鉴权、错误处理等漏洞
- 不一把倒出标准答案
- 紧扣你「服务员」类比里的缺口
- 像助教追问,不像代写讲解
- 你答错也先追问缺口,不直接给标准答案段落
- 未知标出;不编造负责人、日期或指标
包 C——求职(绝不编造经历)
硬规则:模型可以整理、说清你做过的事,但不得编造雇主、职位、数据或技能。缺数字就写 [METRIC],由你对照真实记录自己填。
C1 — 按 JD 改写简历要点
目标:用职位描述的语言改写我的要点:不增加我没写过的经历。
职位描述(关键句):---
[粘贴]
---
我的要点:---
[粘贴]
---
格式:表 旧要点 | 改写 | 复用的 JD 短语 | 是否新增事实?(必须为否)
边界:简历里没有的就删掉或标 [NEED EVIDENCE]。
C2 — 求职信骨架
目标:只凭证据写三成段求职信。
职位:[title] @ [company]
为何选他们:[我自己查到的 2 个事实]
为何是我:[仅来自简历的 3 个证据]
格式:≤250 词;不要「我充满热情」空话。
边界:不编奖项、数据或共同人脉。
C3 — 面试故事(STAR)
目标:把经历收成 STAR(情境、任务、行动、结果)。
经历笔记:---
[粘贴]
---
格式:4 段短文并标注;结果只能用我给的数字或 [METRIC]。
边界:不要英雄式夸张。
C4 — 练习题 + 反馈
目标:针对 [岗位] 问我一道常见题。等我答完。再按清晰度/证据/长度打 1–5 分,并给一句更紧的开场。
边界:不要替我编造更漂亮的履历。
C5 — 风险扫描
目标:扫描简历/求职信里无法核实的说法、含糊动词、AI 腔口号。
格式:问题 | 为何危险 | 更稳妥的改法方向。
边界:宁可诚实的小成果,不要唬人的假大。
实例
填好的提示词
目标:按 JD 改写要点,不增加经历。
JD 关键句:「SQL 看板;向干系人同步;支持 A/B 测试」
我的要点:「为增长做每周 Looker 视图;周五给 PM 发笔记;帮分析师建实验表」
格式:旧 | 改写 | JD 短语 | 是否新增?(否)
边界:不新增雇主或数据。
像样的输出大概长这样
- Looker 映射到看板,却不自称「包办全部 SQL」
- 「是否新增」一列全是否
- 缺数据就留 [METRIC],绝不瞎写 +40%
- 旧 | 改写 | JD 短语 对照清楚
- 面试官追问时你对得上自己的经历
包 C 实战拒绝:模型一编数据,立刻粘贴下面这段。
实例
填好的提示词
停下。「转化 +40%」不在我的要点里。
重做表格。「是否新增」必须仍为否。
缺口用 [METRIC]。不要自己填。
边界:不编造;未知标 [?];发送前停下。
完成检查:一分钟内能判过/不过。
像样的输出大概长这样
- 编造的「转化 +40%」从表格消失
- 「是否新增」仍全为否
- 缺口用 [METRIC] 露出,方便你回去补
- 重做后的表面试站得住
- 拒绝话术可原样再贴进对话
- 面试官追问时你能指回自己的表,不是聊天记录
- 未知标出;不编造负责人、日期或指标
包 D——办公周常(邮件 · 纪要 · 周报)
可搭配更深页面:邮件打磨、会议纪要→行动、干系人周报。本包是周一到周五的起步套件。
D1 — 会议→行动
目标:从记录列出决定、行动、未决问题。
格式:## 决定(仅明确写出的)## 行动(负责人+日期或待定)## 未决 ## 不明确
边界:不编负责人或截止日期。
记录:---
[粘贴]
---
D2 — 周报
目标:给 [上司/客户] 写可扫读的同步。
格式:状态色(绿/黄/红)+ 3 进展 + 3 风险 + 3 求助。≤150 词。
边界:不要空话开头。数字只能用我提供的。
D3 — 催办但不啰嗦
目标:就 [未完成项] 跟进,已过 [N] 天。
语气:礼貌、直接、一个请求。
格式:≤80 词;含主题行。
边界:不要编造并不存在的紧急感。
D4 — 议程
目标:把目标收成 30 分钟会议议程。
目标:[成果]
格式:5 行计时 | 负责人 | 需要拍板?(Y/N)
边界:留缓冲;不要塞 12 个议题。
D5 — 澄清含糊需求
目标:把含糊请求变成我可以回问的 5 个澄清问题。
请求:---
[粘贴]
---
格式:编号问题;每个用一句话就能答。
边界:在我回答前不要开始干活。
实例
填好的提示词
目标:从记录列出决定、行动、未决。
记录:「发版推迟;Maya 周五确认;两个外包预算已批;文档负责人不明。」
边界:不编负责人或截止日期。
完成检查:一分钟内能判过/不过。
像样的输出大概长这样
- 「两个外包预算已批」进决定
- Maya 周五确认动,负责人真实
- 文档负责人不明 → 留在未决,不瞎编死线
- 三类分栏一眼能扫
- 会后五分钟能填完
- 未决项带着「谁、何时」,不丢进虚空
- 未知标出;不编造负责人、日期或指标
包 D 第二个实例:D3(催办但不啰嗦)。
实例
填好的提示词
目标:就 API 冻结日跟进,已过 4 天。
语气:礼貌、直接、一个请求。
格式:≤80 词;含主题。
边界:不编造并不存在的紧急感。
背景:上次请 Maya 周五确认;今天周二;冻结日仍待定。
像样的输出大概长这样
- 全文 ≤80 词,只有一个请求
- 没有「全公司都阻塞了」式假紧急
- 主题行点出 API 冻结日
- 背景时间线(上周请 Maya、今天周二)交代清楚
- 礼貌、直接,像真人跟进
- 收件人能在手机上 10 秒看懂要回什么
今晚填括号
收藏夹里的空包帮不了任何人。睡前用真实事实填完一套。
红旗
整包复制不改姓名日期,发出去的是别人的场景;先替换占位再跑。
保存这张卡
记下:你改过的那一版填空(含日期),别只收藏空模板。
开始前
先选场景包,再粘贴;边界与诚信句保持置顶。
本周怎么用
本周从包里抽一条,把姓名日期场景换成今天真实材料再跑;占位未换不发出。
常见坑
| 坑 | 改法 |
|---|---|
| 空模板收藏了就算会 | 填进真实事实与边界再存 |
| 求职包忘了拒编造 | 硬规则每次置顶 |
| 学习包被当成代写 | 诚信句置顶,拒写可上交正文 |
现在就练一次
今晚就练:从包里抽一条,把姓名日期场景换成今天真实材料再跑。
闭环
写下一行:下次占位未换就发出,我拒绝。
一句话带走
先替换姓名日期场景占位;再跑、再发出。
下一步
← Back to home · Learn · Tools · Compare · Start