Imported from uni-medical/gmai-web (.claude/skills/gmai-member/SKILL.md). Install upstream with npx skills add uni-medical/gmai-web --skill gmai-member. Copyright stays with the author.

GMAI Lab — Team Member Management Skill

When to Use

Trigger phrases:

• "Add a new member" / "添加新成员"
• "Update member X" / "更新成员信息"
• "Remove member X" / "删除成员"
• "Move X to past members" / "把X移到过往成员"
• "Add photo for X" / "添加头像"

Repository Quick Reference

_data/team.yml                     Master team data (pi, co_pis, members, alumni)
pages/team/{slug}.md               EN profile page
pages/zh/team/{slug}.md            ZH profile page
assets/images/team/                Team photos (firstname_lastname.jpg)
_layouts/profile.html              Profile page layout (reads from team.yml)
pages/team.md                      EN team listing (shows Scholar/GitHub/Website only)
pages/zh/team.md                   ZH team listing

Data Structure Hierarchy

pi:           # Single PI (Junjun He) — displayed as main card
co_pis:       # Array — displayed side-by-side with PI in .team-pis flex container
members:      # Array — displayed in .team-grid (4 per row)
alumni:       # Array — displayed in Past Members table (sorted newest-first)

Profile lookup order (_layouts/profile.html): pi.slug → co_pis[].slug → members[].slug (alumni use member slug too)

# Local dev
export GEM_HOME="$HOME/.local/share/gem/ruby/3.2.0"
export PATH="$GEM_HOME/bin:$PATH"
bundle exec jekyll serve --baseurl "" --port 4000
# Deploy: git push origin main

Operations

Operation: ADD

Step 1 — Collect Information (Interactive)

Ask the user for each field. Mark required fields with *.

 1. * Full name (English)?        e.g. "Wentao Li"
 2. * Chinese name?               e.g. "李文涛"
 3. * Role?                       [Researcher / Research Intern]
 4.   If Research Intern: University/institution?
 5. * Research focus (English)?    e.g. "Computational pathology, agent systems"
 6.   Research focus (Chinese)?    (auto-translate if user skips)
 7.   Photo file path?            Full path or "none" for initials placeholder
 8.   Initials?                   2 chars, auto-derived from name if not given
 9.   Links — provide any you have:
      - Google Scholar URL?
      - GitHub URL?
      - Personal website URL?
      - OpenReview URL?
      - Semantic Scholar URL?
      - DBLP URL?
      - HuggingFace URL?
      - LinkedIn URL?
      - ResearchGate URL?
10.   Brief bio (Chinese, 1-2 sentences)?  For ZH profile page
11.   Position in list?           "after [name]" or "at end"

Step 1b — CRITICAL: Identity & Name Verification

Same-name collision warning: Before proceeding, verify the person's identity:

• If a Google Scholar or other link is provided, CONFIRM it belongs to the correct person
• Search the link and check: university, co-authors, research topics — do they match?
• NEVER assume a Scholar link is correct just because the name matches — there are many researchers with identical names (e.g. "Yirong Chen" at SCUT vs Stanford)
• If uncertain, ask the user to confirm before using any links

Chinese name verification:

• Double-check Chinese characters with the user — similar-sounding characters are easily confused
• Examples of past mistakes: 翊 vs 毅, 隆 vs 龙
• If the user provides a Chinese name verbally/phonetically, confirm the exact characters

Step 2 — Generate Slug

slug = lowercase(firstname-lastname)   # e.g. "wentao-li"

Verify slug is unique: grep "slug.*{slug}" _data/team.yml

Step 3 — Process Photo

If photo provided:

# Copy to standard location
cp {user_path} assets/images/team/{firstname}_{lastname}.{ext}

# Check dimensions — crop to square if needed
python3 -c "
from PIL import Image
img = Image.open('assets/images/team/{file}')
w, h = img.size
if w != h:
    s = min(w, h)
    left = (w - s) // 2
    top = (h - s) // 2
    img.crop((left, top, left+s, top+s)).save('assets/images/team/{file}', quality=90)
    print(f'Cropped {w}x{h} → {s}x{s}')
"

Step 4 — Add to _data/team.yml

Insert at specified position in members: array:

  - name:     "FirstName LastName (中文名)"
    slug:     "firstname-lastname"
    role:     "Research Intern, University Name"
    role_zh:  "科研实习生，大学名称"
    focus:    "Research focus English"
    focus_zh: "研究方向中文"
    photo:    "assets/images/team/firstname_lastname.jpg"
    initials: "FL"
    links:
      scholar: ""
      github:  ""
      website: ""

Only include link fields that have values. Add extended links (openreview, semantic_scholar, etc.) only if provided.

Ordering rules:

• Researchers go before # ── Research Interns comment
• Research Interns go after it
• Grid is 4 per row — position affects visual layout

Step 5 — Create EN Profile Page

Write pages/team/{slug}.md:

---
layout:      profile
title:       "FirstName LastName"
permalink:   /team/{slug}/
lang:        "en"
member_slug: "{slug}"
---

Step 6 — Create ZH Profile Page

Write pages/zh/team/{slug}.md:

---
layout:      profile
title:       "中文名 — GMAI Lab"
permalink:   /zh/team/{slug}/
lang:        "zh"
member_slug: "{slug}"
---

中文名来自XX大学，是 GMAI 实验室科研实习生。研究方向包括...

Step 7 — Verify

bundle exec jekyll build --baseurl ""
curl -s -o /dev/null -w "%{http_code}" http://localhost:4000/team/{slug}/
curl -s -o /dev/null -w "%{http_code}" http://localhost:4000/zh/team/{slug}/

Operation: UPDATE

Step 1 — Identify Member

Ask: "Which member? (name or slug)"

Find in team.yml: grep -n "slug.*{input}" _data/team.yml

Step 2 — Ask What to Update

What to update? [photo / focus / role / links / bio / all]

Step 3 — Apply Changes

Use Edit tool to modify the specific fields in _data/team.yml.

If updating photo: copy new file to assets/images/team/, crop if needed, update photo: field.

If updating bio: edit the ZH profile page markdown content.

Operation: REMOVE

1. Remove member entry from members: in _data/team.yml
2. Ask: "Delete profile pages too? [y/n]"
   • If yes: delete pages/team/{slug}.md and pages/zh/team/{slug}.md
3. Ask: "Delete photo? [y/n]"
   • If yes: delete from assets/images/team/

Operation: MOVE_TO_PAST

Step 1 — Collect Info

1. When did they join?              e.g. "2024.06"
2. When did they leave?             e.g. "2025.12"
3. Current position/affiliation?    e.g. "PhD Student, The Hong Kong Polytechnic University"
4. Research focus (short)?          e.g. "AI for Science, Proteomics"

Period format rules:

• Use YYYY.MM format (e.g. "2024.06"), NOT "2024.01" unless truly January
• Verify dates with user — don't guess start dates
• Past mistake: "2024.01 – 2025.12" was incorrect, actual was "2025.06 – 2025.12"

Step 2 — Execute

1. Remove from members: array in _data/team.yml
2. Add to alumni: section with correct schema:

alumni:
  - name:     "Name (中文名)"
    slug:     "firstname-lastname"
    period:   "2024.06 – 2025.12"
    position: "PhD Student, University Name"
    focus:    "Research focus keywords"

Critical: Alumni sorting — alumni must be sorted by end date, newest first. Insert the new entry at the correct position (not just appended to the end).

3. Profile pages are kept (still accessible via direct URL)
4. The Past Members table in pages/team.md / pages/zh/team.md auto-renders from alumni data
5. Name display uses {{ a.name | replace: " (", "<br>(" }} for EN/CN line break

Operation: ADD CO-PI

For leadership-level members displayed alongside the main PI:

1. Add to co_pis: array in _data/team.yml (between pi: and members:)
2. Same field structure as pi: (name, slug, role, title, bio, photo, links)
3. Displayed side-by-side with PI in .team-pis flex container
4. Profile lookup in _layouts/profile.html checks co_pis after pi
5. Create profile pages same as regular members

Operation: FILL PROFILE CONTENT

Fill rich Markdown content for a member's profile subpage. Profile pages support {{ content }} rendering — anything written below the front matter in pages/team/{slug}.md becomes the profile body.

When to Use

Trigger phrases:

• "Fill profile for X" / "写X的简介"
• "Add bio content for X" / "填充X的个人页面"
• After creating a member's profile pages (ADD operation Steps 5-6)

Step 1 — Collect Information (Interactive)

Ask the user for or derive from context:

 1. * Current position + institution?        e.g. "PhD Student, Monash University"
 2.   Supervisor(s)?                         e.g. "Prof. Jianfei Cai"
 3.   Previous positions (chronological)?    e.g. "Senior R&D Engineer at Baidu → Researcher at Shanghai AI Lab"
 4.   Role in GMAI team?                     e.g. "Core member, principal architect of SA-Med2D-20M"
 5. * Top 3-5 research areas?                With key project names
 6. * Selected publications (4-8 items)?     Title, venue, year, author role, one-line description
 7. * Career path summary?                   For Career Trajectory section
 8.   Brief bio in Chinese?                  Auto-translate from EN if not provided

Step 2 — Determine Template

Regular Members use the 4-section template:

PI / Co-PI use the 6-section template:

Step 3 — Write EN Profile Content

Edit pages/team/{slug}.md — append Markdown below the front matter ---:

Example (Regular Member — Jin Ye):

---
layout:      profile
title:       "Jin Ye"
permalink:   /team/jin-ye/
lang:        "en"
member_slug: "jin-ye"
---

## About

Jin Ye is a PhD student at **Monash University**... Before starting his PhD,
he was a full-time researcher at **Shanghai AI Laboratory**...

## Research Interests

- **Large-scale medical segmentation datasets**: SA-Med2D-20M (4.6M images / 19.7M masks)
- **SAM adaptation for medicine**: SAM-Med2D, SAM-Med3D, SAM-Med3D-MoE
- ...

## Selected Publications

- **SA-Med2D-20M** (arXiv 2023) — *First Author*. The largest 2D medical segmentation dataset...
- **SAM-Med3D** (ECCV 2024 Workshop Oral; IEEE TNNLS 2025) — Extension of SAM to 3D...
- ...

## Career Trajectory

**Baidu** (Senior R&D Engineer) → **Shanghai AI Laboratory** (Researcher, GMAI Team) → **Monash University** (PhD Student)

Step 4 — Write ZH Profile Content

Edit pages/zh/team/{slug}.md — same structure with Chinese text:

---
layout:      profile
title:       "叶锦 — GMAI Lab"
permalink:   /zh/team/jin-ye/
lang:        "zh"
member_slug: "jin-ye"
---

## 关于

叶锦目前是**蒙纳士大学**...博士研究生...

## 研究方向

- **大规模医学分割数据集**：SA-Med2D-20M（460万张图像 / 1970万个标注掩码）
- ...

## 代表性论文

- **SA-Med2D-20M**（arXiv 2023）— *第一作者*。当时最大的2D医学分割数据集...
- ...

## 职业轨迹

**百度**（高级研发工程师）→ **上海人工智能实验室**（研究员，GMAI 团队）→ **蒙纳士大学**（博士研究生）

Content Guidelines

• Length hierarchy: Regular member content MUST be shorter than PI's. PI gets Highlights + Academic Overview + Early Works; members do not.
• About: 1-2 paragraphs. Bold institution names. Mention GMAI role/contribution.
• Research Interests: 3-5 bullet points. Bold the area label, include key project names in parentheses.
• Selected Publications: 4-8 items. Format: **Paper Name** (Venue Year) — *Role*. One-line description. Include GitHub stars if notable.
• Career Trajectory: Single line with bold institution names and arrow → separators. End with a brief summary phrase after —.
• Always fill both EN and ZH pages. If user doesn't provide Chinese text, auto-translate.
• Use **bold** for institution names and project names.
• Use *italic* for roles like "First Author", "2nd Author".

Reference Implementations

• PI template: pages/team/junjun-he.md (EN), pages/zh/team/junjun-he.md (ZH)
• Regular member template: pages/team/jin-ye.md (EN), pages/zh/team/jin-ye.md (ZH)

Data Schema Reference

Team listing page shows (3 links only):

• scholar → "Scholar"
• github → "GitHub"
• website → "Website" / "主页"

Profile page shows ALL links:

• scholar, github, website
• openreview, semantic_scholar, dblp
• huggingface, linkedin, researchgate
• email, cv, twitter

Photo requirements:

• Location: assets/images/team/{firstname}_{lastname}.{jpg|png}
• Recommended: square crop, ≤500KB
• If no photo: set photo: "" and provide initials: "XX"
• CSS renders as 96px circle with object-fit: cover

Checklist Before Commit

• _data/team.yml entry added/updated
• EN profile page exists: pages/team/{slug}.md
• ZH profile page exists: pages/zh/team/{slug}.md
• Profile content filled (if not empty placeholder): About, Research Interests, Selected Publications, Career Trajectory
• Both EN and ZH profile content provided (auto-translate ZH if needed)
• Member content is shorter than PI content
• Photo copied and cropped (if provided)
• Links: only non-empty fields included
• Jekyll build passes
• Both EN and ZH team pages render correctly