Person Intelligence

Airtable SLT (if SLT member -- Gary, Adam, Ashleigh, Michael, Anish, Kevin):

python3.12 ~/.claude/local-plugins/nsls-personal-toolkit/skills/person-intelligence/scripts/fetch_airtable_slt.py "{name}" > /tmp/person-intel-slt.json

Airtable People Ops (if NSLS employee):

python3.12 ~/.claude/local-plugins/nsls-personal-toolkit/skills/person-intelligence/scripts/fetch_airtable_people_ops.py "{name}" > /tmp/person-intel-people-ops.json

Existing Obsidian profiles (vault at $OBSIDIAN_VAULT_PATH):

• 30-people/{Name}.md (display name with spaces)
• 10-slt/members/{slug}.md (lowercase hyphenated, e.g., gary-tuerack.md)
• 20-projects/board-intelligence/members/{Name}.md (display name with spaces)

Step 3: Fetch and summarize meetings

If no email provided or Fathom returns 0 matches, skip this step. The synthesizer works with Airtable data alone.

Fetch transcripts and summarize each:

python3.12 ~/.claude/local-plugins/nsls-personal-toolkit/skills/person-intelligence/scripts/fetch_fathom_1on1s.py \
  --email {email} --fetch-all > /tmp/person-intel-meetings.jsonl

# Summarize each meeting (one Claude API call per meeting)
while IFS= read -r line; do
  echo "$line" | python3.12 ~/.claude/local-plugins/nsls-personal-toolkit/skills/person-intelligence/scripts/summarize_meeting.py
done < /tmp/person-intel-meetings.jsonl > /tmp/person-intel-summaries.jsonl

For weekly updates: Add --after {last-synthesized date} to only fetch new meetings.

Performance: First runs with 20+ meetings take 20+ API calls. This is expected. Subsequent weekly runs are fast (1-3 new meetings).

Step 4: Infer project connections

Assemble goals, actions, and topics from all sources into a JSON object and pipe to the inference engine:

python3.12 ~/.claude/local-plugins/nsls-personal-toolkit/skills/person-intelligence/scripts/infer_projects.py < /tmp/person-intel-data.json

Input format: {"goals": [...], "actions": [...], "topics": [...], "person_name": "..."}

• Goals: from Airtable L1/L2/LOP goals
• Actions: from Airtable Meeting Actions
• Topics: from meeting summaries "Topics Discussed" sections

Step 5: Synthesize profile

Assemble ALL collected data into a single JSON payload and synthesize:

python3.12 ~/.claude/local-plugins/nsls-personal-toolkit/skills/person-intelligence/scripts/synthesize_profile.py < /tmp/person-intel-combined.json > /tmp/person-intel-profile.md

Input format:

{
  "person_name": "...",
  "meeting_summaries": [...],
  "airtable_slt": {...},
  "airtable_people_ops": {...},
  "existing_profile": "...",
  "existing_board_profile": "...",
  "existing_slt_profile": "...",
  "projects": {...}
}

All fields except person_name are nullable -- the synthesizer handles any combination.

Step 6: Write to Obsidian

Write the synthesized profile to $OBSIDIAN_VAULT_PATH/30-people/{Name}.md.

If the person is an SLT member, also update $OBSIDIAN_VAULT_PATH/10-slt/members/{slug}.md with coaching feedback patterns (speaking %, contribution quality trends, start/stop recommendations over time).

If the person is a board member, do NOT overwrite $OBSIDIAN_VAULT_PATH/20-projects/board-intelligence/members/{Name}.md -- that has hand-curated board context. The 30-people/ profile links to it instead.

Step 7: Surface project suggestions

If infer_projects.py returned "suggested" projects (1-2 matches, below the 3-match confirmation threshold), present them:

"I found possible project connections for {name}:

• {project}: mentioned {N} times ({evidence}). Add to profile?"

If confirmed, add to the profile AND update the project's collaborators: frontmatter.

Step 8: Update project collaborators

For confirmed projects, update $OBSIDIAN_VAULT_PATH/20-projects/{project}/{project}.md frontmatter: