product-os: Command for Claude Code

Meeting Prep

Generate a prep brief for an upcoming meeting by pulling context from previous meetings, open actions, quarterly goals, and the person's 1:1 context file.

Instructions

Step 0: Parse arguments and identify the meeting

• The user provides arguments in the format: $ARGUMENTS
• Expected format: {person name} (e.g. {CPO}, Alex, Product small group)
• If no argument is provided:
  1. Use mcp__claude_ai_Google_Calendar__gcal_list_events to get events in the next 24 hours
  2. Filter to meetings that look like 1:1s or recurring syncs (exclude all-hands, large group meetings, focus time)
  3. Present the next 3 candidate meetings using AskUserQuestion and ask the user to confirm which one to prep for
• Once the meeting is identified, determine:
  • Person name — the other attendee (for 1:1s) or the meeting name (for group meetings)
  • Slug — lowercase, hyphenated version (e.g. alex-smith, jane-doe)
• Check the auto-memory (MEMORY.md) to identify the user's initials and their direct reports/manager
• Classify the meeting type:
  • report — if the person is one of the user's direct reports
  • manager — if the person is the user's manager
  • other — any other meeting (peer, exec, group sync)

Step 1: Gather context (run in parallel)

Pull from all available sources. Run independent queries in parallel.

1. Recurring meeting context file — Read team/{initials}/recurring-meeting-context/{slug}.md

   • If the file exists: extract standing topics, development themes, carry-forward items, and relationship notes
   • If the file does not exist: note this and proceed. After the brief is generated, offer to create one

2. Previous meeting notes (Granola) — Use mcp__claude_ai_Granola__list_meetings (with last_30_days or a custom range) to find recent meetings, then filter by this person's name in the title or participants to identify the most recent 1-3 meetings

   • Use mcp__claude_ai_Granola__get_meetings (batch, up to 10 IDs) to retrieve notes and summaries
   • If notes are sparse (just a few words or empty), use mcp__claude_ai_Granola__get_meeting_transcript and extract key points
   • From the most recent meeting, extract:
     • Topics discussed
     • Decisions made
     • Open actions (especially anything assigned to the user or the other person)
     • Unresolved questions or threads