roast-my-pi

One wrong advisor, five years gone.
Data-driven X-ray for academic advisors — see through titles and team glory.

First author on Nature ≠ Author #28 on Nature.

Authorship position, grant ownership, co-author power structure — these three things decide whether an advisor is a real PI or just riding a team ticket.

This tool does one thing: it takes a potential academic advisor's publications, funding, and collaboration network apart, strips away the institutional halo, and tells you what they personally have actually done.

Independence Ratio First / corresponding author vs. middle-author filler 0.30 = the PI threshold

Grant Ownership Their own funding vs. borrowed grant numbers General Program = real PI

Power Structure Lab head vs. peer vs. trainee Who actually owns the project?

---

🚩 Red Flags Cheat Sheet

Save these. Screenshot them. You'll never read a CV the same way.

1. Author position beats journal impact factor

First author on Nature is not the same as author #28. Where you sit on the author list says everything about what you actually did. Position 20+ usually means you handed over some samples — that's it.

Same Nature paper	What it actually means
1st author / corresponding	They drove the work
3rd–5th author	Meaningful contribution
20th+ author	Sent in some samples

2. Independence ratio below 0.15? Not a PI.

Score every paper by the author's role:

Role	Weight
First author	+1.0
Corresponding (last + marked)	+1.0
Co-first author	+0.7
Co-corresponding	+0.5
Middle author	+0

independence_ratio = weighted_score / total_papers

Score	Verdict
> 0.30	Credible independent researcher
0.15 – 0.30	Building independence, not there yet
< 0.15	A bench hand, not a lab head

3. "Last author" ≠ PI

People assume the last name on the list is the PI. Wrong — the corresponding-author marker is what actually tells you.

• Last position + corresponding marker = real PI (weight 1.0)
• Co-corresponding, not last = half the claim (weight 0.5)
• Last position, no corresponding marker = probably just the senior name on the door

4. A grant in the acknowledgments isn't their grant

Seeing a grant number in the footnotes doesn't mean it belongs to them.

Cross-check: pull every paper citing that grant ID and see who's corresponding. If it's always someone else, they're surfing on borrowed money.

The real owner of a grant is the person who keeps showing up as corresponding author on papers funded by it.

5. Only ever held a Youth grant? They're stuck.

In the Chinese NSFC funding system, grant tier maps directly to career stage:

Grant type	Prefix	What it means
Youth Project	8xxxxxx	Entry-level (~210K CNY) — everyone starts here
General Program	3xxxxxx	The minimum bar for an independent PI (~550K)
Excellent Young	2xxxxxx	Elite tier (~2M)
Key Program	7xxxxxx	Senior PI territory (~3M)

Years of papers but never a General Program? That's what Chinese academia calls a "permanent attending" (万年主治) — stuck at the same rank, never independent.

6. Eight years, same boss, still a middle author — that's a ceiling.

• If they keep publishing with the same senior PI and never move up from the middle of the list, they've never carved out their own direction.
• Papers without the lab head? That's the real independence signal.
• The trajectory matters more than the snapshot: climbing, or flatlined?

---

📊 Sample Report

Synthetic case disclaimer: this report is fictional, built from real evaluation patterns. No data refers to any actual person.

Advisor Insight Report

Institution: university-affiliated hospital | Data source: common PubMed patterns

Independence ratio: 0.18 — team dependent, not an independent PI (threshold: 0.30)

Author role breakdown Role Share First author ~5–10% Corresponding < 5% Middle author ~80% Large consortium (>20) ~10%	Journal quality Tier Picture T1 Elite (IF > 15) A handful — always middle author T2 Top (IF 8–15) Several T3–T4 (IF 2–8) The bulk; first-author papers here T5 Low / predatory Yes, some
Grant ownership Type Status NSFC Youth Project 1, completed NSFC General Program Never Institutional grants A couple	Co-author power structure Metric Finding Lab head (Boss) On ~80% of papers, always last This person's position Position 5–8 / 10–15, year after year Papers without Boss Almost zero

Verdict: this researcher has spent years as the workhorse middle author in a senior PI's lab. Plenty of high-IF papers, but almost never in the driver's seat. Independence ratio of 0.18 sits well below the 0.30 bar. Only a Youth grant — never a General Program. Not someone who can independently mentor a PhD student.

---

🚀 Quick Start

# Load the skill in Hermes Agent
skill_view(name='advisor-insight')

# Then follow the guided workflow

You'll need to provide:

Required	Optional
Full name (Chinese + English)	ORCID (gold standard for ID)
Current institution + department	Known collaborators / former students
Research-area keywords	Title or rank

A name alone won't cut it. You need at least institution + department for disambiguation.

---

⚙️ How It Works

Phase 0 ── Disambiguation ──────► Cross-check PubMed, ORCID, Scholar, institution pages
                                   Make sure we've got the right person
    │
Phase 1 ── Publication Audit ───► Pull every PubMed paper, classify author role,
                                   compute independence ratio
    │
Phase 2 ── Grant Analysis ──────► Extract grant IDs from acknowledgments,
                                   trace who actually owns each one
    │
Phase 3 ── Co-author Network ──► Count co-authorships, map power structure
                                   and the lab's food chain
    │
Phase 4 ── Career Trajectory ──► Reconstruct the arc from publication history —
                                   is independence growing or stalled?
    │
Phase 5 ── Synthesis ───────────► Standardized score + plain-English verdict

---

🌏 中文版

点击展开完整中文文档

避坑指南：6 条识破注水导师

每一条都能截图单发。读完你不会再被简历唬住。

1. 看作者位置，别看期刊影响因子

Nature 一作不等于 Nature 第 28 作者。署名位置才是真实贡献度：第 20+ 作者通常只贡献了样本。盯住一作和通讯论文的数量与质量。

2. 独立性比率 < 0.15，基本不是独立 PI

把所有论文按署名角色打分：一作 +1.0，通讯（末位 + 通讯标记）+1.0，共同一作 +0.7，共同通讯 +0.5，中间作者 +0。独立性比率 = 加权得分 / 论文总数。> 0.30 可信的独立研究者；0.15 – 0.30 独立性发展中；< 0.15 团队螺丝钉。

3. 通讯作者才是真 PI，末位作者可能只是挂名

末位 + 通讯标记 = 真 PI（权重 1.0）。共同通讯（非末位）= 半个 PI（权重 0.5）。末位但无通讯标记 = 可能只是资深挂名。

4. 查基金归属，不是看致谢里有没有基金号

交叉验证：搜索所有用了这个基金号的论文，看谁在这些论文里当通讯。如果全是别人通讯的论文在致谢，他只是蹭了课题组成员的基金。

5. 只有青年基金 + 院内课题，大概率万年主治

基金类型	编号前缀	含义
青年项目	8xxxxxx	入门级，约 21 万
面上项目	3xxxxxx	独立 PI 的门槛，约 55 万
优青	2xxxxxx	精英级，约 200 万
重点项目	7xxxxxx	资深 PI，约 300 万

6. 跟了同一个 Boss 八年还在中间位置，就是天花板

看合作网络的权力结构：有没有不带 Boss 独立发的论文？独立性的轨迹比当下状态更重要。

---

这是什么

roast-my-pi（导师锐评） 是一个 Hermes Agent 技能，系统性地客观评估一位潜在学术导师的真实科研画像：剥离机构光环、虚高头衔、团队荣誉，揭示个人的实际独立贡献。

核心原则：数据胜于声誉，证据胜于修辞。

它能告诉你什么

• 这个人是真正的独立 PI，还是平台依赖型贡献者
• 实际发文独立性比率，对比团队挂名比率
• 名下到底有没有独立基金（不是蹭的）
• 合作网络中的真实权力位置（Boss / 平级 / 学生）
• 职业轨迹是上升、停滞、还是已经触顶
• 你加入他的组，竞争力和发展前景如何

快速开始

skill_view(name='advisor-insight')

需要准备：导师全名（中英文）+ 别名、当前机构 + 科室、职称、ORCID（如有）、已知合作者、研究方向关键词。至少需要机构 + 科室才能开始消歧。

---

License

MIT — free to use, modify, and distribute.

Contributing

Contributions welcome — better heuristics, new data sources, additional analysis dimensions. Open an issue or PR.