Lab Roster

Careful information gathering using subagents for each member.

When to use: First-time generation, complex updates, verification needed Output: ROSTER.md + lab_roster.csv

Instructions

Step 1: Determine Mode

Ask the user which mode they prefer, or choose based on context:

• If ROSTER.md doesn't exist or is outdated → Mode 2
• If just need CSV update → Mode 1
• If user wants verification → Mode 2

Step 2a: Quick Generation (Mode 1)

1. Run the roster generation script:

   python3 .claude/skills/lab-roster/scripts/generate_roster_csv.py
   

2. Validate output:

   • Check row count matches expected members
   • Verify currently active members
   • Show summary statistics

3. Present results to user with summary

Step 2b: Comprehensive Generation (Mode 2)

1. Check if ROSTER.md exists

   • If exists, read it to understand current state
   • If not, create it with field descriptions and instructions

2. Get list of all members

   ls _members/*.md | grep -v "friends"
   

3. Launch subagents to gather information

   • Process members in batches of 4-5
   • Each subagent should:
     • Read member file: _members/[name].md
     • Check git history: git log --follow --format="%aI|%s" -- _members/[name].md
     • Check for role transitions in git: git show <commit>:_members/[name].md
     • Check alumni list in team/index.md
     • Infer team from bio keywords (see reference.md)
     • Extract co-advisor from bio text
     • Extract previous and next positions

4. Update ROSTER.md

   • Add entries in the format from templates/roster-entry.md
   • One entry per role (create multiple entries for role transitions)
   • Only fill "Previous" field in first entry
   • Only fill "Next" field in last entry

5. Generate CSV from ROSTER.md

   python3 .claude/skills/lab-roster/scripts/roster_md_to_csv.py
   

6. Validate and present

   • Show summary statistics
   • Highlight any missing data
   • Ask user to review

Field Descriptions

• name: Full name of the lab member
• role: Short code (pi, postdoc, staff, phd, ms, ra, programmer, undergrad, highschool)
• start_date: When they started (YYYY-MM format)
• end_date: When they ended (YYYY-MM format, empty if still active)
• previous_position: Position before joining lab (only in first entry)
• next_position: Where they went after (only in last entry)
• team: Primary team (software_engineering, phenoinformatics, virtual_biology)
• co_advisor: Other PIs they work with

Team Inference Rules

Use these keyword mappings to infer team from bio text:

software_engineering:

• Keywords: SLEAP, DREEM, cloud, infrastructure, plant, root, AWS, full-stack, data pipeline, software engineer, programmer, bioinformatics analyst, computer vision algorithm, pose estimation

phenoinformatics:

• Keywords: phenotyp, behavior, ALS, Alzheimer, disease model, mice, mouse, longitudinal, space, neurodegenerative, multi-animal, tracking pipeline

virtual_biology (highest priority):

• Keywords: VNL, virtual lab, virtual animal, embodied, neuromechanical simulation

Note: Check virtual_biology keywords first, then software_engineering, then phenoinformatics.

Co-Advisor Extraction

Look for these patterns in bio text:

• "co-advised by [Name]"
• "co-supervised by [Name]"
• "jointly with [Name]"
• "joint with [Name]"

Remove markdown links and extract just the name.

Previous/Next Position Extraction

Previous positions - look for:

• "Prior to joining..."
• "previously worked at/as..."
• "received [degree] from..."
• Education background

Next positions - look in team/index.md alumni section:

• Format: **Next:** [position]

Validation Checklist

Before presenting results:

• All active members included
• Role transitions properly split into multiple entries
• Start dates seem reasonable
• Teams assigned for most members
• No duplicate entries
• CSV has same data as ROSTER.md (if both exist)
• Summary statistics make sense

Output Format

Always show:

✓ Generated roster with [N] entries
✓ Saved to: lab_roster.csv

Summary:
  Total entries: [N]
  Unique members: [N]
  Currently active: [N]
  Alumni entries: [N]
  Team distribution:
    - software_engineering: [N]
    - phenoinformatics: [N]
    - virtual_biology: [N]

Error Handling

If script fails:

1. Check Python version (needs Python 3.6+)
2. Verify git is available
3. Check that _members/ directory exists
4. Look for malformed member files

If subagents miss information:

1. Review their search patterns
2. Check if member files follow expected format
3. Manually verify edge cases

Notes

• Exclude members with role "friends"
• Multiple role entries for same person are normal (e.g., undergrad → MS → PhD)
• Start dates are best guesses from git history
• Some fields may be empty - that's okay
• ROSTER.md is human-editable - CSV regenerates from it

See Also

• reference.md - Detailed documentation on member formats and inference patterns
• examples.md - Example usage scenarios
• templates/roster-entry.md - Entry format template