Client meeting prep briefs for wealth advisors

Build me an agent workflow that runs every weekday at 6:30am America/New_York and files a one-page client meeting prep brief as a Google Doc for each client review on my calendar that day. I am a wealth advisor and this runs before my workday starts.

Trigger: cron, every weekday (Mon-Fri) at 6:30am ET.

Inputs the agent should accept as configuration:

- work_calendar_id: the Google Calendar to read (default: primary).

- client_meeting_title_prefix: optional string like "Review" or "Annual Review" used to identify client meetings by event title.

- client_attendee_domains: optional list of email domains that mark an event as a client meeting (use either the prefix, the domains, or both — if both are set, an event matching either rule counts).

- holdings_sheet_id and holdings_range: the Google Sheets spreadsheet that maps client name to holdings. Assume columns: client_name, ticker, security_type ("stock" or "mutual_fund"), shares. Mutual fund tickers in the sheet may or may not include the M: prefix — normalize before calling YCharts.

- output_drive_folder_id: optional folder where the briefs should live (otherwise drop in root).

Steps:

1. Call google-calendar List Events on work_calendar_id for today's date range in America/New_York (timeMin = today 00:00 ET, timeMax = today 23:59 ET, singleEvents = true). Filter to events whose title starts with client_meeting_title_prefix OR whose attendees include an address whose domain is in client_attendee_domains. Skip declined events and events you organized but have no external attendees on.

2. For each matched client meeting, derive the client name. Prefer an explicit "Client: <name>" line in the event description; fall back to the non-internal attendee's display name, then to the event title with the prefix stripped.

3. Call google-sheets Get Values on holdings_sheet_id/holdings_range and filter rows where client_name matches (case-insensitive, trimmed). If no rows match, still produce a brief with the meeting context but a note that no holdings were found.

4. Split holdings by security_type. For stocks/ETFs, batch tickers (up to 100) into a single ycharts Get Company Data Points call with calculation codes: price, day_price_pct_change, ytd_total_return, dividend_yield. For mutual funds, prepend M: to each ticker and batch into a single ycharts Get Mutual Fund Data Points call with the same calculation codes. Handle YCharts response envelope (top-level meta + response keyed by symbol) and surface any per-symbol errors gracefully without aborting the whole brief.

5. Compute, per holding: market_value = shares * price; day_change_value = market_value * (day_price_pct_change / 100). Sum to portfolio_value. Rank holdings by day_change_value to identify the top contributors and top detractors. Flag any holding whose absolute day move is more than 2% as "news-worthy."

6. Draft a one-page brief with these sections: header with client name, meeting time (in ET), and attendees; Portfolio snapshot (total value, day change in $ and %, YTD return weighted by market value, weighted dividend yield); Biggest contributors today (top 3 with $ impact); Biggest detractors today (bottom 3 with $ impact); Notable moves (any "news-worthy" flagged holdings); Suggested talking points (2-3 concise bullets the advisor can riff on, grounded in the data — for example a position that has driven YTD performance, a position whose dividend yield has shifted, a fund whose YTD diverges from the rest of the book).

7. Call google-docs Create Document with title "<Client name> — Prep <YYYY-MM-DD HH:MM ET>" then google-docs Batch Update Document to insert the brief content with simple structure (Heading 1 for the client name, Heading 2 for each section, bullets where appropriate). If output_drive_folder_id is set, move the doc into that folder.

8. After all meetings are processed, write a short run summary to the workflow log: how many meetings matched, how many briefs were created, any clients skipped for missing holdings, and any YCharts symbols that returned errors.

Guardrails: never make up prices or returns — if YCharts has no value for a symbol, mark it as "data unavailable" in the brief instead of guessing. Keep each brief to roughly one page. Use a consistent professional tone, not casual. Do not include client personal data beyond what is on the calendar and in the holdings sheet.