Meeting Cadence Optimizer

Meeting Cadence Optimizer

Find the meeting load where the user's days actually go well -- grounded in their own logged data, not vibes. This skill reads back their Evening Debrief records (day rating, meeting count, cadence feedback) and Morning Check-In records (energy), then computes the relationship between meeting density and day quality. Let the data do the convincing.

Prerequisites

• Fulcra CLI authenticated (uv tool run fulcra-api auth login).
• Run with uv run --python 3.12.
• Meaningful output needs history: this works once there are several evening-debrief entries with meeting_count and day_rating. The skill reports a confidence level so you know how much to trust it.

Analyze (read-only)

uv run --python 3.12 ~/.claude/skills/meeting-cadence-optimizer/scripts/cadence.py analyze --days 30

Returns, computed from the user's own records:

• data_points and a confidence gate: insufficient (<7 days), low (7-13), medium (14-29), high (30+).
• avg_meeting_count, avg_day_rating
• day_rating_by_meeting_count -- the average day rating at each meeting count (the core table)
• best_rated_meeting_count -- the count with the highest average rating (the "sweet spot")
• meeting_count_vs_rating_correlation -- Pearson r (negative = more meetings, worse days)
• too_many_feedback_count -- how often they flagged "too many"

How to present it

Match your confidence to the data, and only state numbers the script returned -- never invent a correlation:

• insufficient: "I need about a week of evening debriefs before I can find your pattern -- keep wrapping up your days and I'll have something soon." Don't push numbers.
• low: "Early signal, low confidence -- here's what I'm seeing so far..."
• medium / high: Lead with the sweet spot and the table, e.g. "Your days average 7.4 with 3-4 meetings but drop to 5.1 at 6+. Your sweet spot looks like 4. You've flagged 'too many' 5 times this month." Then offer one or two concrete moves: cap daily meetings, add buffers after long ones, protect a focus block.

Keep recommendations actionable and few. The goal is a decision, not a dashboard.

Save the analysis (optional)

To record the analysis as a Fulcra annotation (useful for a weekly cadence so trends are themselves trackable):

uv run --python 3.12 .../cadence.py save --days 30 --dry-run   # preview
uv run --python 3.12 .../cadence.py save --days 30             # write

Writes a "Cadence Analysis" moment annotation and reports verified_matches. If there's no data yet, save skips the write rather than recording an empty analysis.

Weekly use

This is a natural Sunday-evening or Monday-morning review. If the user wants it automated, it can be scheduled to run weekly and surface the result in their next morning briefing.

Privacy

Meeting patterns and day ratings are personal. Don't surface them publicly. Never print API tokens.

Where this fits

The analytical capstone of the Fulcra concierge reflection loop (shared lib in ~/.fulcra-concierge/lib). It consumes what evening-debrief (meeting count + day rating) and subjective-checkin (energy) record -- so its quality grows directly with how consistently those two are used.