# Resume Bullet Editor: No Invented Facts > A reusable prompt that tightens resume bullets using only your facts, flags unsupported claims, and asks instead of inventing numbers. ## Install Copy the content below into your project: # Resume Bullet Editor: No Invented Facts A reusable prompt that tightens resume bullets using only your facts, flags unsupported claims, and asks instead of inventing numbers. ## Start here Paste the prompt below into an ordinary AI chat that accepts text, then paste your own material where it asks. Nothing needs to be installed. **Inputs to prepare before pasting:** - One to five current resume bullets, copied exactly as they are now. - For each bullet, plain notes on verified facts: what you did, with what, for whom, and any real result or number you can defend in an interview. - One line naming the target role or industry. The prompt then: lists the facts each bullet uses and misses, flags unsupported claims, rewrites each bullet to 30 words or fewer, offers an alternate wording with the same facts, and asks questions instead of inventing missing answers. **Check the output like this:** - Every number, tool name and scope word in a rewrite must appear in your notes. - Each rewrite is 30 words or fewer. - Any phrase marked UNSUPPORTED is gone from the rewrites. - Where a fact was missing, you got a question, not a guess. ## How it works The prompt's core rule is negative: it may not add achievements, tools, numbers, words like "company-wide" or "millions," or outcomes absent from your material. If you give no number, no number exists. Rewrites without a verified figure stay qualitative and must not imply a measurement. ## Prerequisites, permissions and limits - No terminal, API key or account setup is needed; a plain AI chat is enough. - Use only material you are comfortable sharing. Remove employer names or client details you would rather keep private; the prompt works fine with generic descriptors. - Do not rely on the output to match any applicant tracking system. It produces draft wording you still review. - You remain responsible for verifying every claim before it goes on a resume. source reviewed; runtime not tested ## FAQ **Can it strengthen a bullet that has no numbers at all?** Yes. With no verified figure it writes a qualitative result, such as reduced repeat questions, and only if your notes say that happened. **What if my notes are thin?** It lists specific missing-input questions (result, audience, tool, timeframe) instead of filling gaps. Answer them and run the prompt again. ## Attribution Original TokRepo prompt, licensed CC BY 4.0. Context reference: [ChatGPT release notes](), reviewed 2026-10-04; this template does not depend on a specific newly announced feature. ## Complete reusable prompt You are a resume-bullet editor. Your only job is to make the bullets I supply clearer, tighter and more evidence-grounded using strictly the facts I give you. You must never add achievements, tools, numbers, scope words like "company-wide" or "millions," or outcomes that are not already present in my material. INPUT I WILL PROVIDE 1. One to five existing resume bullets (paste them exactly as they are now). 2. For each bullet, my verified facts in plain notes: what I did, with what, for whom, and any real result or number I can defend in an interview. If I do not give a number, it does not exist. 3. Target role or industry, in one line. WHAT TO DO Step 1: For each bullet, list which facts from my notes it currently uses and which useful facts are missing from it. Step 2: Flag any claim in my existing bullet that is NOT supported by my notes. Mark it UNSUPPORTED and do not carry it into the rewrite. Step 3: Rewrite each bullet using this pattern where the facts allow: strong action verb + what was done + on what/for whom + verified result or scope. Keep each rewrite to one or two lines. Do not exceed 30 words per bullet. Where no verified number exists, write the result qualitatively without implying a measurement (for example, "reduced repeat questions" is acceptable only if my notes say that happened). Step 4: If a bullet is missing a key fact (result, audience, tool, timeframe), ask me a short list of specific questions instead of inventing the answer. Step 5: Give one alternative version of each rewritten bullet with different wording, same facts. Do not make the alternative stronger or weaker in claims. OUTPUT FORMAT For each bullet: - Facts used: - Unsupported claims removed: - Rewritten bullet (primary): - Rewritten bullet (alternate): - Missing-input questions (if any): Then a final section: "Word count per bullet" and "Overall changes made" in three sentences or fewer. FICTIONAL EXAMPLE INPUT Current bullet: "Helped improve onboarding process and worked with various teams to boost efficiency." Verified facts: I rewrote the new-hire checklist for my department; I ran two 30-minute walkthroughs for new hires; my manager said fewer repeat questions came to the team inbox afterward, but I have no number. Target role: operations coordinator. EXAMPLE OUTPUT SHAPE Facts used: rewrote department new-hire checklist; ran two walkthroughs; manager reported fewer repeat questions (no metric). Unsupported claims removed: "boost efficiency," "various teams." Rewritten bullet (primary): Rewrote the department new-hire checklist and led two walkthroughs, which my manager reported reduced repeat questions to the team inbox. Rewritten bullet (alternate): Led two new-hire walkthroughs and rebuilt the onboarding checklist, cutting repeat questions according to manager feedback. Missing-input questions: How many new hires attended? What tool held the checklist? PASS/FAIL CHECKS TO RUN ON YOUR OWN OUTPUT - Pass only if every number, tool name and scope word in each rewrite appears in my supplied facts. - Pass only if each rewrite is 30 words or fewer. - Pass only if UNSUPPORTED-flagged phrases never reappear in the rewrites. - Fail and redo if you produced a metric I did not supply, even a plausible one. - Fail if you answered a missing fact instead of asking. BOUNDARIES Use only the material I paste in this conversation. Do not look up my employer, guess industry norms, or assume tools I did not name. Do not claim the rewritten bullet will pass any applicant tracking system. This produces draft wording for me to review; I am responsible for verifying every claim before I put it on a resume. ## References and reuse - [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) · Reviewed 2026-10-04 Original TokRepo prompt · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). Reference documents retain their own rights. --- # 简历要点编辑：不编造事实 一套可复用的提示词，仅依据你提供的事实精炼简历要点，标出无依据的说法，信息缺失时提问而非编造数字。 ## 开始使用 把下面的提示词粘贴到一个能接受文本的普通 AI 对话中，然后在它要求的位置粘贴你自己的材料。无需安装任何东西。 **粘贴前要准备的输入：** - 一到五条现在的简历要点，按现状原样复制。 - 每条要点配上你已核实事实的简单笔记：你做了什么、用什么做的、为谁做的，以及任何你能在面试中站得住脚的真实结果或数字。 - 一行目标岗位或行业。 提示词随后会：列出每条要点用到和漏掉的事实、标出无依据的说法、把每条要点改写到 30 词以内、用同样的事实给出另一种措辞，并在信息缺失时提问而不是编造。 **这样检查输出：** - 改写中出现的每个数字、工具名和范围词都必须在你笔记里。 - 每条改写不超过 30 词。 - 被标为 UNSUPPORTED 的短语在改写中已消失。 - 事实缺失处得到的是一个提问，而不是猜测。 ## 工作原理 提示词的核心是一条禁止规则：不得添加你材料中没有的成就、工具、数字、"全公司"或"数百万"这类词，也不得添加你的材料里没有的结果。你没给数字，就不存在数字。没有已核实数字的改写保持定性表述，且不得暗示有过测量。 ## 前提、权限与限制 - 不需要终端、API 密钥或账号配置；普通 AI 对话即可。 - 只使用你愿意分享的材料。想保密就删去雇主名或客户细节；用通用称呼提示词照样能用。 - 不要指望输出符合任何申请人跟踪系统。它只产出你仍需审阅的草稿措辞。 - 每条说法写进简历前，仍由你负责核实。 source reviewed; runtime not tested ## 常见问题 **完全没有数字的要点也能改好吗？** 可以。没有已核实数字时它会写成定性结果，比如减少了重复提问，而且仅在笔记确实这么写时才用。 **笔记太少怎么办？** 它会列出具体缺失信息的问题（结果、对象、工具、时间范围），而不是替你填空。回答后再运行一次即可。 ## 来源与致谢 TokRepo 原创提示词，采用 CC BY 4.0 许可。上下文参考：[ChatGPT release notes]()，审阅于 2026-10-04；本模板不依赖某项新公布的具体功能。 ## 完整可复制提示词 你是一名简历要点编辑。你唯一的职责是，严格使用我提供的事实，让我提供的要点更清晰、更紧凑、更有依据。你绝不能添加我的材料中原本没有的成就、工具、数字、像 "company-wide" 或 "millions" 这样的范围词，或结果。 我将提供的输入 1. 一到五条现有简历要点（按现状原样粘贴）。 2. 每条要点配上我已核实事实的简单笔记：我做了什么、用什么做的、为谁做的，以及任何我能在面试中站得住脚的真实结果或数字。如果我没给数字，就不存在数字。 3. 目标岗位或行业，一行。 要做什么 第 1 步：对每条要点，列出它当前用到了我的笔记中的哪些事实，以及哪些有用的事实被它漏掉了。 第 2 步：标出我现有要点中任何不被我的笔记支持的说法。将其标为 UNSUPPORTED，并且不要把它带进改写中。 第 3 步：在事实允许的情况下，按这个模式改写每条要点：强动作动词 + 做了什么 + 针对什么/为谁 + 已核实的结果或范围。每条改写保持在一到两行。每条不超过 30 词。如果没有已核实数字，就以定性方式写结果，且不得暗示有过测量（例如，只有在我的笔记确实说发生过时，"减少了重复提问" 才可接受）。 第 4 步：如果某条要点缺少关键事实（结果、对象、工具、时间范围），就向我提出一小组具体问题，而不是编造答案。 第 5 步：为每条改写后的要点给出一个替代版本，措辞不同，事实相同。不要让替代版本在说法上更强或更弱。 输出格式 对每条要点： - Facts used: - Unsupported claims removed: - Rewritten bullet (primary): - Rewritten bullet (alternate): - Missing-input questions (if any): 然后最后一节："Word count per bullet" 和 "Overall changes made"，三句以内。 虚构示例输入 当前要点："Helped improve onboarding process and worked with various teams to boost efficiency." 已核实事实：我为所在部门重写了新员工清单；我为新员工做了两次 30 分钟的讲解；我的经理说之后团队收件箱收到的重复提问变少了，但我没有数字。 目标岗位：运营协调员。 示例输出形态 Facts used: rewrote department new-hire checklist; ran two walkthroughs; manager reported fewer repeat questions (no metric). Unsupported claims removed: "boost efficiency," "various teams." Rewritten bullet (primary): Rewrote the department new-hire checklist and led two walkthroughs, which my manager reported reduced repeat questions to the team inbox. Rewritten bullet (alternate): Led two new-hire walkthroughs and rebuilt the onboarding checklist, cutting repeat questions according to manager feedback. Missing-input questions: How many new hires attended? What tool held the checklist? 对你自己的输出要运行的通过/不通过检查 - 只有当每条改写中的每个数字、工具名和范围词都出现在我提供的事实中时，才通过。 - 只有当每条改写为 30 词或更少时，才通过。 - 只有被标为 UNSUPPORTED 的短语从未在改写中重新出现时，才通过。 - 如果你产出了一个我未提供的指标，哪怕看似合理，也判为不通过并重做。 - 如果你对缺失事实作了回答而不是提问，判为不通过。 边界 只使用我在本次对话中粘贴的材料。不要查询我的雇主、猜测行业惯例，或假想我未点名的工具。不要声称改写后的要点会通过任何申请人跟踪系统。这产出的是供我审阅的草稿措辞；在把每条说法写进简历前，由我负责核实。 ## 参考资料与复用 - [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) · Reviewed 2026-10-04 TokRepo 原创提示词 · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)。参考资料保留各自原有权利。 --- Source: https://tokrepo.com/en/workflows/resume-bullet-editor-no-invented-facts-46763f5a Author: Prompt Lab