Back to Details

admissions-cv-writing

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Admissions CV / Resume Writing

入学CV/简历撰写

Use When

适用场景

The user wants to create, rewrite, polish, or export an English graduate admissions CV or resume from their education, research, internship, project, activity, award, and skills background.
当用户希望基于自身的教育、科研、实习、项目、活动、获奖及技能背景创建、改写、润色或导出英文研究生入学CV/简历时使用。

Workflow

工作流程

Use the user's language for all conversation, questions, explanations, and review notes, but keep the actual CV output in English unless the user explicitly asks otherwise.
所有对话、提问、解释和审核备注都使用用户的语言，但实际CV输出默认保留英文，除非用户明确要求其他语言。

Phase 1: Collect Information

第一阶段：信息收集

Load ONLY references/info-requirements.md, then guide the user using that checklist and the material-mining prompts.
  • Show the checklist so the user can provide everything at once if they prefer.
  • After receiving input, identify what is already covered and ask follow-up questions only for missing required categories or experiences with thin detail.
  • Collect by module in order: education -> experiences (priority: research -> internship -> project -> campus activity -> extracurricular activity) -> awards -> skills.
  • Mine each experience for bullet-point material: what the user did, which methods/tools they used, and what outcomes followed.
  • Use careful inference only as a questioning strategy. You may suggest likely details in question form, but you must keep only what the user confirms and discard what they reject or skip.
  • Optional categories should be asked only when useful; they are not mandatory.
⛔ Do not proceed to Phase 2 until every required category in references/info-requirements.md has been addressed. A category is addressed only when the user has either provided all required sub-fields or explicitly confirmed they have nothing for that category. See the Validation Rules section in that reference.
仅加载 references/info-requirements.md，参照该清单和信息挖掘提示引导用户提供材料。
  • 向用户展示清单，方便偏好一次性提供所有材料的用户使用。
  • 收到用户输入后，确认已覆盖的内容，仅针对缺失的必填类别或信息过于简略的经历追问补充。
  • 按模块顺序收集：教育经历 -> 经历（优先级：科研 -> 实习 -> 项目 -> 校园活动 -> 课外活动）-> 获奖情况 -> 技能。
  • 挖掘每段经历的要点素材：用户做了什么、使用了哪些方法/工具、取得了什么成果。
  • 仅可将合理推断作为提问策略：你可以用提问的方式给出可能的细节建议，但必须仅保留用户确认的内容，丢弃用户拒绝或跳过的内容。
  • 可选类别仅在有价值时询问，非必填。
⛔ 在 references/info-requirements.md 中的所有必填类别都处理完毕前，不得进入第二阶段。仅当用户提供了某个类别的所有必填子字段，或明确确认该类别无内容可填时，才算该类别处理完成，详见该参考文档中的验证规则部分。

Phase 2: Confirm Information and Judge Sufficiency

第二阶段：信息确认与充足度判断

  1. Summarize the collected information by section and ask the user to confirm accuracy and missing items.
  2. Evaluate whether the material is sufficient using the criteria in references/info-requirements.md.
  3. Tell the user clearly whether the material is sufficient or thin, and explain why.
  4. If the material is thin, ask whether the user can add more detail such as actions, tools, methods, or measurable outcomes. If they cannot, explain that you can carefully infer draft bullet points for review.
    • If the user chooses to supplement, return to the relevant module in Phase 1 and then repeat Phase 2.
    • If the user chooses to skip supplementation, continue to Phase 3 Branch B.
⛔ Do not proceed to Phase 3 until the user explicitly confirms the summary and, when the material is thin, explicitly chooses whether to supplement or skip.
  1. 按模块汇总收集到的信息，请用户确认准确性和缺失项。
  2. 参照 references/info-requirements.md 中的标准评估材料是否充足。
  3. 明确告知用户材料是否充足或信息过薄，并解释原因。
  4. 如果材料过薄，询问用户是否可以补充更多细节，比如动作、工具、方法或可量化成果。如果用户无法补充，说明你会谨慎推断要点草稿供用户审核。
    • 如果用户选择补充，返回第一阶段对应的模块收集信息，之后重复第二阶段流程。
    • 如果用户选择跳过补充，进入第三阶段B分支。
⛔ 在用户明确确认汇总信息，且当材料过薄时用户明确选择补充或跳过之前，不得进入第三阶段。

Phase 3: Produce the CV (Review Version)

第三阶段：生成CV（审核版）

Load references/writing-instructions.md and references/cv-format-example.md, then follow their drafting rules. The output at this phase is a clean review version without HTML tags, intended for human reading and confirmation.
加载 references/writing-instructions.md 和 references/cv-format-example.md，遵循其中的起草规则。本阶段输出为无HTML标签的干净审核版本，供人工阅读和确认。

Branch A: Sufficient Material

分支A：材料充足

Generate the review version:
  1. User-language summary — a concise section-by-section synopsis for quick fact-checking.
  2. English full text — the complete CV content in plain text (no HTML 
    div
    tags), following the structure and writing rules from the references.
生成审核版：
  1. 用户母语摘要 — 按模块整理的精简摘要，用于快速事实核对。
  2. 英文全文 — 纯文本格式的完整CV内容（无HTML 
    div
    标签），遵循参考文档中的结构和写作规则。

Branch B: Thin Material

分支B：材料过薄

Step 1: Expand bullet points for review
Load references/bullet-expansion-guide.md and generate at least 3 bullet points per experience from the available material.
  • Output each bullet in English plus a short explanation in the user's working language when that helps review.
  • Clearly mark inferred content on the explanation line with 
    [Inferred]
    .
  • Append a follow-up checklist of details that would still strengthen each experience.
⛔ Do not proceed to Step 2 until the user explicitly confirms or revises the expanded bullet points.
Step 2: Produce the review version
After user confirmation, generate the review version:
  1. User-language summary — same as Branch A.
  2. English full text — include user-approved content; omit rejected inferred content entirely; strip all markers (
    [Inferred]
    , 
    [Needs Detail]
    ).
  3. AI Notes — append after the English full text, listing remaining gaps and omitted content.
步骤1：扩充要点供审核
加载 references/bullet-expansion-guide.md，基于现有材料为每段经历生成至少3个要点。
  • 每个要点输出英文版本，必要时附加用户使用语言的简短解释方便审核。
  • 在解释行用 
    [Inferred]
    明确标记推断生成的内容。
  • 附加后续可优化每段经历的细节清单。
⛔ 在用户明确确认或修改完扩充的要点之前，不得进入步骤2。
步骤2：生成审核版
用户确认后，生成审核版：
  1. 用户母语摘要 — 同分支A。
  2. 英文全文 — 包含用户确认的内容，完全删除用户拒绝的推断内容，移除所有标记（
    [Inferred]
    、
    [Needs Detail]
    ）。
  3. AI Notes — 附加在英文全文之后，列出剩余的信息缺口和省略的内容。

Phase 4: Quality Check and User Confirmation

第四阶段：质量检查与用户确认

Load references/quality-checklist.md and run the checklist against the review version (user-language summary + English full text).
  • Group results by severity: Error / Warning / Pass.
  • If there are Error-level issues, fix them and present the revised review version.
  • After all errors are resolved, ask the user to confirm the final content.
⛔ Do not proceed to Phase 5 until the user explicitly confirms the review version.
加载 references/quality-checklist.md，对照审核版本（用户母语摘要 + 英文全文）执行检查。
  • 按严重程度分组展示结果：错误 / 警告 / 通过。
  • 如果存在错误级问题，修复后展示修订后的审核版本。
  • 所有错误解决后，请用户确认最终内容。
⛔ 在用户明确确认审核版本之前，不得进入第五阶段。

Phase 5: Export

第五阶段：导出

After user confirmation, ask whether the user wants a PDF. If yes:
  1. Generate the tagged English Markdown from the confirmed English full text:
    • Add HTML 
      div
      layout tags following the patterns in references/cv-format-example.md.
    • The tagged version is for rendering only — it must not alter, add, or remove any content from the confirmed English text.
  2. Save the tagged Markdown as a 
    .md
    file at the user-specified path. If none is given, save to the current working directory with the filename 
    CV_<Full-Name>_<YYYYMMDD>.md
    (e.g., 
    CV_Zhang_Yuhan_20260403.md
    ).
  3. Consistency check: compare the saved tagged Markdown (with tags stripped) against the confirmed English full text.
    • This is a content-level comparison; minor differences in symbols or formatting are acceptable.
    • If a content discrepancy is found, fix the file and re-check.
  4. Export to PDF (automatically initializes the environment on first run):
    bash
    bash scripts/export-pdf/run.sh <input.md> <output.pdf>
    • Default font mode is 
      auto
      : prefer bundled fonts when the platform allows packaged font files, otherwise fall back to similar local system fonts.
    • If the platform forbids packaged font files, use 
      --font-source local-only
      to force local-font rendering.
    • If bundled fonts are required for brand consistency, use 
      --font-source bundled-only
      so the export fails fast when font assets are unavailable.
All script paths are relative to the 
admissions-cv-writing/
skill directory.
Prerequisites: PDF export requires Python 3 and an internet connection (for first-time dependency installation). The setup script creates an isolated venv and installs 
weasyprint
and 
markdown
automatically. No global packages are modified. If bundled fonts are unavailable, PDF export still works by falling back to locally installed fonts with a similar style, but line breaks and spacing may vary slightly across platforms.
用户确认后，询问用户是否需要PDF版本。如果需要：
  1. 基于确认后的英文全文生成带标签的英文Markdown：
    • 参照 references/cv-format-example.md 中的格式添加HTML 
      div
      布局标签。
    • 带标签版本仅用于渲染，不得修改、新增或删除已确认英文文本中的任何内容。
  2. 将带标签的Markdown保存到用户指定路径的 
    .md
    文件。如果未指定路径，保存到当前工作目录，文件名为 
    CV_<Full-Name>_<YYYYMMDD>.md
    （例如 
    CV_Zhang_Yuhan_20260403.md
    ）。
  3. 一致性检查：将保存的带标签Markdown（移除标签后）与确认的英文全文对比。
    • 这是内容层级的对比，符号或格式的细微差异可接受。
    • 如果发现内容不一致，修复文件后重新检查。
  4. 导出为PDF（首次运行时自动初始化环境）：
    bash
    bash scripts/export-pdf/run.sh <input.md> <output.pdf>
    • 默认字体模式为 
      auto
      ：平台允许打包字体文件时优先使用内置字体，否则 fallback 到相似的本地系统字体。
    • 如果平台禁止打包字体文件，使用 
      --font-source local-only
      强制使用本地字体渲染。
    • 如果品牌一致性要求必须使用内置字体，使用 
      --font-source bundled-only
      ，当字体资源不可用时导出会快速失败。
所有脚本路径都相对于 
admissions-cv-writing/
技能目录。
前置要求：PDF导出需要Python 3和网络连接（用于首次依赖安装）。设置脚本会创建隔离的venv环境，自动安装 
weasyprint
和 
markdown
，不会修改全局包。如果内置字体不可用，PDF导出仍可通过 fallback 到风格相似的本地安装字体正常工作，但不同平台的换行和间距可能略有差异。

Output Format

输出格式

  • Review version (shown to user): user-language summary + clean English full text, no HTML tags.
  • Tagged Markdown (written to file): English full text with HTML 
    div
    layout tags, for PDF rendering only. Never shown directly to the user.
  • Branch B review: additionally append an 
    AI Notes
    section after the English full text.
  • Quality check: group results as Error / Warning / Pass.
  • 审核版（展示给用户）：用户母语摘要 + 干净的英文全文，无HTML标签。
  • 带标签的Markdown（写入文件）：带有HTML 
    div
    布局标签的英文全文，仅用于PDF渲染，不得直接展示给用户。
  • B分支审核版：额外在英文全文后附加 
    AI Notes
    模块。
  • 质量检查：按错误 / 警告 / 通过分组展示结果。