Calendar Profile Generation
Input Data Structure
The session data (Calendar Raw Data) contains:
- events: Array with attendees, duration, recurrence info
- collaborators: Pre-scored by meeting patterns
- summary.eventsByDayOfWeek: Meeting distribution by day
Key Collaborators
Use the collaborators array but apply judgment:
- Focus on people with recurring 1:1s or consistent meeting patterns
- Note external vs. internal (different email domain = external)
- Ignore people who only appear in large group meetings
- Weight 1:1s heavily—they're the strongest collaboration signal
Schedule Patterns
Identify from the data:
- Heavy meeting days (from eventsByDayOfWeek)
- Whether mornings or afternoons tend busier
- Any meeting-free patterns (e.g., "No meetings Friday afternoons")
Profile Output
Write markdown to Connection Profiles.
Structure:
Last analyzed: [ISO datetime]
## Timezone
[User's default timezone]
## Key Collaborators
[Names with relationship context—"weekly 1:1", "external client", etc.]
## Schedule Patterns
[Busy days, typical rhythms]
## Meeting Types
[What kinds dominate—1:1s, standups, client calls, etc.]
DO NOT include raw counts. Describe patterns narratively.