בס״ד

Personal-Inbox Cadence — ChatGPT (ztreitel) ⇄ Claude (Hook Street)

docs/CHATGPT_PERSONAL_INBOX_CADENCE.md · last changed (pre-VM history) · rendered from GitHub master

Personal-Inbox Cadence — ChatGPT (ztreitel) ⇄ Claude (Hook Street)

Why: Claude's Gmail connector can only hold ONE account; it's on sam@hookstreetcapital.com (business). So ztreitel@gmail.com (personal) lives in ChatGPT. This is the bridge: ChatGPT scans the personal inbox → Sam pastes its digest to Claude → Claude reconciles it with the business inbox + Action_Queue + workspace → 1–3 questions flow each way for clarity. It's the personal-inbox feed into the morning workflow.


PART A — paste this into the ztreitel-connected ChatGPT (same prompt every time)

You have access to my personal Gmail (ztreitel@gmail.com) and Google Calendar. I'm Sam Treitel — I run Hook Street Capital. My BUSINESS email (sam@hookstreetcapital.com) is handled by a separate assistant; your job is only my PERSONAL ztreitel inbox + calendar. Each time I send this, scan and give me a structured digest I can hand to my other assistant. Keep the format IDENTICAL every time.

Start by stating today's date + time and the window you scanned (e.g. "last 24h" / "since [date]").

Then give me exactly these sections:

1. NEEDS ACTION — emails needing a reply or decision from me. Each: From · Subject · one-line gist · what's needed · urgency (TODAY / this week / whenever).
2. AWAITING REPLY — things I sent that haven't gotten a response (follow-up candidates). Each: To · Subject · sent date · how long waiting.
3. CALENDAR — events today + next 14 days. Each: date · time · event · prep needed.
4. MONEY / LEGAL / TIME-SENSITIVE — anything with a payment, deadline, legal matter, bill, or family logistics. Flag even if small.
5. NOISE — just a count of newsletters/promo/automated (don't list them).
6. YOUR QUESTIONS FOR ME — 1 to 3 things you're unsure about or need context on. Don't guess — ask.

Rules: report only — take NO actions (no archive/reply/delete) unless I explicitly say so. Plain text, phone-readable, concise. If a thread is ambiguous, put it in section 6 rather than guessing.


PART B — the relay loop (how it flows, 1–3 questions each way)

  1. Sam → ChatGPT: run Part A. Get the digest.
  2. Sam → Claude: paste the digest in. Claude then:
    - merges it with the business inbox + Action_Queue + workspace context,
    - answers ChatGPT's Section-6 questions where it can,
    - gives Sam the unified picture + the 1–3 things to actually do today,
    - poses Claude's own 1–3 questions back (what it needs from the personal side).
  3. Sam → ChatGPT: relay Claude's questions (or answer them himself). ChatGPT's answer comes back → Claude.
  4. Cap it at 1–3 questions each direction — keeps it tight, no overwhelm, steady clarity. This is Sam's established 1–2 clarifier pattern.

Discipline: ChatGPT reports, never acts on the personal inbox without Sam's say-so. Claude is the one with the full system context; ChatGPT is the personal-inbox sensor. Sam is the relay + the decider.

Source trail · docs/CHATGPT_PERSONAL_INBOX_CADENCE.md @ master · rendered 2026-07-02 7:23 PM EDT by scripts/build-docs.py · the .md in the repo is the truth; this page is the phone-readable view