A career advisor's brain, distilled into a prompt skill

What it does

It’s an Agent Skills package (plain markdown + YAML frontmatter) that turns any skills-compatible agent — Claude Code, Codex, Cursor, and 50+ others — into a Zhang Xuefeng-style advisor for gaokao志愿, 考研, and career planning. The authors did deep research: 5 books, 15+ media interviews, 30+ first-hand quotes, 11 key decision records, and a full life timeline, then compressed it into 5 mental models, 8 decision heuristics, and a documented “expression DNA” (short punchy sentences, survival-and-salary vocabulary).

The interesting bit

The pitch is explicit: this is not ChatGPT wearing a Zhang Xuefeng mask. Every response is supposed to apply his actual cognitive frameworks — 「社会筛子论」, 「就业倒推法」, 「阶层现实主义」 — to your question rather than replaying famous quotes. It’s persona engineering as decision methodology, which is a genuinely different approach from most persona prompts.

Key highlights

• 5 named mental models, each traced to a source (e.g. 「选择 > 努力」 appears in two of his books)
• 8 decision heuristics like 「中位数原则」 (judge a major by where the middle 50% of graduates end up) and 「家庭背景分流」 (same question, different answer depending on family resources)
• Works across runtimes via the open Agent Skills protocol; the content is just markdown, so you can paste SKILL.md into any chat
• Example transcripts in examples/ show the persona handling AI-era career questions, not just classic gaokao ones
• Generated by the authors’ own 女娲.skill (nuwa-skill) skill-generator, which is the meta-joke here

Caveats

• The advice persona is deliberately blunt and opinionated (“家里没钱，别谈理想”) — it’s a stylistic feature, but expect strong takes, not balanced counseling
• The README doesn’t show how the skill is validated; quality depends on how well the underlying prompt file carries the models
• It’s a persona/decision-style artifact, not a data source — the employment claims in sample answers are rhetorical, not cited statistics

Verdict

Worth a look if you build agent personas or want a well-structured example of encoding a decision framework into a skill. If you want actual college admissions data or neutral advice, this is the wrong sieve.