Free Layer · Free Forever免费层 · 永久免费
Ten Claude Code skills that walk you from choosing a subject to defending your paper in English at the Tsinghua finals — grounded in this archive and the whitepaper, never inventing a winner, judge, or rule.
十个 Claude Code 技能，陪你从选定学科一路走到清华总决赛的全英文答辩—— 所有内容都植根于本档案与白皮书，绝不杜撰任何获奖者、评委或规则。
These are skills for Claude Code: small, invokable assistants you call with
a slash command (e.g. /yau-topic-finder). Each one knows the
丘成桐中学科学奖 deeply — its six subjects, its all-English defense, its judges' tastes,
and 丘成桐's "自己出题自己做" originality ethic — because each distills the local winners
archive (393 entries, 2020–2025) and the whitepaper into its own reference library. They
score topics as ranges, not promises; they refuse to fabricate; and they hand off to one
another in a logical order.
它们是面向 Claude Code 的技能：可用斜杠命令唤起的小型助手（例如 /yau-topic-finder）。
每一个都深谙丘成桐中学科学奖——六大学科、全英文答辩、评委的偏好，以及丘成桐"自己出题自己做"的原创精神——
因为它们各自把本地获奖档案（393 条，2020–2025）与白皮书提炼成了自己的参考库。它们以区间而非承诺来给课题打分；
拒绝凭空编造；并按合理的次序彼此交接。
The Pipeline技能流程
The family mirrors the Yau journey. Start at orientation, move through topic and writing, and finish at the English defense. Two skills run alongside the whole thing.
这套技能映射了完整的丘奖历程：从选向起步，历经选题与写作，终于全英文答辩。另有两个技能贯穿始终、随时相伴。
yau-subject-navigator — pick 1 of 6 subjects六选一，确定学科
yau-topic-finder — discover or score a topic发掘或评分一个课题yau-mentor-finder — find an advisor寻找指导老师
yau-research-plan-drafter — subject-fit outline契合学科的提纲
yau-paper-writeryau-data-analysis-tutoryau-ai-compliance
yau-submission-reviewer — pre-submission materials review提交前材料审查
yau-defense-coach — English finals全英文总决赛
Running alongside, any time: yau-pathway-navigator (rounds, the three tracks,
deadlines, awards). yau-ai-compliance also feeds the defense, since judges
probe AI use directly.
全程相伴、随时可用：yau-pathway-navigator（赛程轮次、三条赛道、截止日期与奖项）。
yau-ai-compliance 也服务于答辩，因为评委会直接追问 AI 的使用情况。
The Ten十个技能
Each card shows what the skill does, the stage it serves, and an example invocation. All are bilingual (English / 中文) and run offline from the local archive.
每张卡片标明该技能的作用、所属阶段，以及一条调用示例。它们全部支持中英双语，并基于本地档案离线运行。
Helps you decide which of the six subjects — 数学 / 物理 / 化学 / 生物 / 计算机 / 经济金融建模 — fits your interests and skills, and what each subject's panel actually values.帮你判断六大学科——数学 / 物理 / 化学 / 生物 / 计算机 / 经济金融建模——中哪一门最契合你的兴趣与能力，以及各学科评委真正看重什么。
The entry point. Discovers topics from your profile, or scores a topic you have — weighting originality and your own ownership above scale, with range-based confidence and no medal promises.起点所在。它从你的背景出发发掘课题，或为你已有的课题打分——把原创性与你本人的主导权置于规模之上，以区间给出信心，绝不承诺奖项。
Helps you ethically find and approach an advisor (drafts only — you send), and enforces the Yau boundary: the student leads, the mentor assists, never authors.帮你以正当方式寻找并联系指导老师（只起草，发送由你完成），并守住丘奖的边界：学生主导，导师辅助，绝不代笔。
Structures a paper outline that fits your subject — a math proof, an experimental study, a CS systems paper, and an econ model each need a different skeleton — then critiques it for rigor.为你的学科量身搭建论文提纲——数学证明、实验研究、计算机系统论文、经济模型各需不同的骨架——再从严谨性上加以审视。
Coaches you to write the 论文 in your own voice — subject norms, the English-quality requirement, citation honesty — so it survives the defense. Never ghost-writes.指导你用自己的笔触写好论文——学科规范、英文质量要求、引用诚信——让它经得起答辩。绝不代笔。
Right-sizes your analysis per subject and flags the landmines judges hammer — correlation-as-causation in econ, no validation set in bio/CS, weak baselines, training-accuracy overreach.按学科为你的分析定好分量，并标出评委最爱穷追的雷区——经济类的相关当因果、生物/计算机缺验证集、基线过弱、用训练准确率夸大成绩。
Navigates the Yau award's implicit AI norms — there is no public ruleset — covering disclosure, per-subject etiquette (math's suspicion vs CS's expectation), and the judge AI Q&A.带你摸清丘奖隐性的 AI 规范——并无公开成文规则——涵盖如何披露、各学科的分寸（数学的警惕与计算机的预期之别），以及评委关于 AI 的问答。
Prepares you for the all-English finals defense at Tsinghua: presentation structure, per-subject question drills, live mock defense, whiteboard readiness, and English-speaking tactics.为清华总决赛的全英文答辩做准备：演讲结构、分学科问题演练、实时模拟答辩、白板应对，以及英文口语技巧。
Explains the rounds (报名与提交 → 材料审核 → 分赛区评审 → 公示 → 总决赛), the three tracks (mainland / Asia / North America), regional divisions, awards, and how to plan a realistic year.讲清赛程轮次（报名与提交 → 材料审核 → 分赛区评审 → 公示 → 总决赛）、三条赛道（内地 / 亚洲 / 北美）、地区分区、奖项设置，以及如何规划切实可行的一整年。
Runs the committee's own materials review before you press submit: eligibility, PDF structure, the 500–1500-character acknowledgements page, the updated integrity declaration, the plagiarism report against the final version, AI disclosure with chat logs — sorted by disqualification risk, with a dated countdown. See the Submission Guide.在你点提交之前，先按组委会的材料审核口径过一遍：资格、PDF 结构、500–1500 字致谢页、更新版诚信声明、与最终版对应的查重报告、AI 披露与聊天记录——按取消资格风险排序，并给出倒排日期。另见提交材料指南。
Getting Started快速上手
These skills run inside Claude Code
and OpenAI's Codex CLI alike —
both read the same open
tian2-skills marketplace
on GitHub, where the Yau family ships as the yau plugin. Two commands on either
platform and you have all ten, updated whenever the marketplace is.
这些技能同时运行于 Claude Code 与
OpenAI 的 Codex CLI——两者读取的是
GitHub 上同一个开放技能市场
tian2-skills，丘奖技能家族在其中以
yau 插件发布。任一平台输入两条命令即可装齐十个技能，市场更新时随之更新。
/plugin marketplace add tian2-ai/tian2-skills./plugin install yau@tian2-skills. The marketplace splits skills by competition — install only what you need, so skill descriptions don't crowd out your working context./yau-topic-finder — optionally with arguments. Installed skills carry the yau: namespace, so they never clash with same-named skills from other sources./plugin marketplace add tian2-ai/tian2-skills。/plugin install yau@tian2-skills。市场按竞赛拆分插件——只装你需要的那个，技能描述才不会挤占真正干活的上下文。/yau-topic-finder——可附带参数。装入的技能带 yau: 命名空间，不会与其他来源的同名技能冲突。Inside any Claude Code session:在任意 Claude Code 会话中：
/plugin marketplace add tian2-ai/tian2-skills /plugin install yau@tian2-skills
Then just start with a slash command:然后，直接用一条斜杠命令开始：
/yau-subject-navigator # not sure which subject? /yau-topic-finder score "my idea" # already have a topic? /yau-submission-reviewer report.pdf # September 15 approaching? /yau-defense-coach --mode mock # finals coming up?
In a terminal with Codex installed:在装有 Codex 的终端里：
codex plugin marketplace add tian2-ai/tian2-skills codex plugin add yau@tian2-skills
Inside a Codex session, type /skills to browse what's installed, mention a skill as $yau-topic-finder to invoke it explicitly — or just describe your task and Codex picks the matching skill itself.在 Codex 会话中，输入 /skills 浏览已装技能，用 $yau-topic-finder 的写法显式调用——或者直接描述你的任务，Codex 会自动匹配对应技能。
Every skill defaults to bilingual output (English + 中文); pass --lang en or
--lang zh to pick one. They draw only on this archive and the whitepaper, so they
work offline and never fabricate a winner, judge, paper, or rule. On AI use, the award
published an official
ruleset on 2026-07-08 — the skills cite it directly (the allowed and prohibited lists, the
four-element disclosure, chat-log submission), and treat the whitepaper's judge-level etiquette
as a supplement to the rules, never a substitute. See the
Submission Guide for the full picture.
每个技能默认输出中英双语；加上 --lang en 或 --lang zh 即可只选其一。
它们只取材于本档案与白皮书，因此可离线运行，绝不杜撰任何获奖者、评委、论文或规则。关于 AI 使用，
丘奖已于 2026 年 7 月 8 日发布官方规则——
这些技能直接引用官方条文（允许与禁止清单、致谢页四要素披露、聊天记录备查），白皮书中的评委礼仪
只作规则之外的补充，绝不替代规则。完整口径见提交材料指南。
These ten skills are the free layer, and they stay free. If you want the layer above — a delivered, committee-grade preflight report on your actual materials, season-long audits, and human review — that is YauLab Pro.
这十个技能是免费层，并且保持免费。如果你需要更上面那一层——对你的真实材料出具组委会 口径的预检报告、全赛季审计与人工复核——那是 YauLab Pro 专业版。