Real Agent

Claude System Prompt for Email Triage

At 7:15am I open Claude on my phone and type one word: triage. Forty seconds later, my Gmail is sorted into a P0/P1/P2 to-do list with time blocks — and the prompt below is what's doing the work.

This page is the prompt itself, lifted from Issue #001 and packaged for copy-and-use. The full build (stack decision, day-by-day failure log, the math) lives there. If you just want the prompt — here it is.

Answer Summary

  • Drives a Gmail triage agent on Claude.ai (Pro plan) using the Gmail Connector — no code, no API key.
  • Reads the last 24 hours of email, sorts into 4 buckets, returns a P0/P1/P2 to-do list with source links.
  • Setup ~20 minutes. Cost $20/month (Claude Pro). Daily runtime ~40 seconds, manual trigger from the phone.
  • Tested for two weeks; saves ~25 minutes of inbox staring per morning.
  • Five [bracketed] sections to customize — role, manager, projects, focus, skip-senders.

What this prompt does

  • Sorts your Gmail inbox into four buckets: must handle today / follow up this week / FYI / skip.
  • Returns a prioritized to-do list with 🔴 P0 (urgent + important), 🟡 P1 (important), 🟢 P2 (optional).
  • Each P0/P1 item includes the email subject, the specific action, a time window, and a time estimate.
  • Ends with a Sources: block — every priority claim links straight back to the original Gmail thread.

Who it's for

  • Anyone who starts the day in Gmail and loses the first hour to inbox triage.
  • Roles where email actually decides the day: analysts, managers, ops, founders, recruiters, support leads.
  • Not for you if your inbox is already nearly empty, or if you don't have Claude Pro + the Gmail Connector.

The full prompt

Your task: triage the user's Gmail inbox and produce a prioritized to-do
list for today, based on email content.

## Execution

1. Pull recent emails
   - Use Gmail tool `search_threads` to search for emails in the last
     24 hours (query: `newer_than:1d` or `is:unread newer_than:2d`).
   - If many unread, also run `is:important newer_than:3d` to catch
     read-but-important threads.
   - For ambiguous emails, use `get_thread` to fetch full content
     before classifying.

2. Categorize each email into one of four buckets, with a one-line
   summary of sender, subject, and core ask:
   - Must handle today (clear deadline, boss/client follow-up,
     interview/meeting confirmation)
   - Follow up this week
   - FYI only (newsletters, notifications, receipts)
   - Skip / archive (promotional, low-value subscriptions, spam)

3. Generate the prioritized to-do list:
   - 🔴 P0: urgent AND important (must be done today)
   - 🟡 P1: important but not urgent (advance today)
   - 🟢 P2: optional / low priority

   For each item include: source email subject, specific action,
   time window (morning / afternoon / evening), and time estimate.

4. Output format: Markdown.
   Start with a one-line summary (e.g., "3 must-dos, 5 follow-ups").
   Then the categorized email summary, then the to-do list.
   End with a "Sources:" section linking each referenced thread as
   `[subject](https://mail.google.com/mail/u/0/#inbox/THREAD_ID)`.

## Constraints
- Do NOT auto-reply or delete any email. Triage and recommend only.
- If the Gmail Connector is unauthorized or unreachable, say so clearly
  and tell the user to re-authorize.
- If there are fewer than 3 emails for the day, say "inbox is quiet"
  and give optional items only. Do not invent tasks to fill space.
- Be concise. No long explanations.

## My context (CUSTOMIZE THIS SECTION)
- Role: [your role, e.g. "Industrial Engineer at Texas Instruments,
  transitioning to Business Planning in April"]
- Manager: [name + email]
- Active projects: [project names — boost any email about these to P0/P1]
- Current focus: [what you're heads-down on this month]
- Always skip senders: [newsletter@..., notifications@..., etc.]

How to customize it

Five spots have [brackets] — those are the only edits that matter. The structure is universal; the customization is where the prompt earns its keep.

  1. [your role, ...] — your job title and current focus. Be specific. "Senior data analyst at a fintech, owning the weekly KPI report" beats "data analyst."
  2. [Manager: name + email] — your manager's email. The agent uses this to detect "your manager replied" and boost priority.
  3. [Active projects] — current project names. Any email mentioning these gets nudged into P0/P1.
  4. [Current focus] — what you're heads-down on this month. Helps the agent pick the right "important" over the wrong "important."
  5. [Always skip senders] — emails or domains you ignore. These cannot land in P0, no matter what subject they use.

Example output

A real morning's output from this prompt:

# Daily Triage — Monday, May 11
3 must-do, 2 follow-ups, 3 quick wins. 13 promotional + 11 system noise filtered.

## 🔴 URGENT — handle in next 2 hours
- **[P0] Submit AIGC Challenge entry** — deadline 5/15 23:59 CDT.
  Action: finalize submission, push to portal. ~60min. Block at 2pm.

## 🟡 IMPORTANT — handle today
- **[P1] Amazon: 3 packages "Delivery attempted"** — WeDoi / SUMK / Lumlazni.
  Action: reschedule via Amazon, choose locker pickup. ~10min. 3:30pm.
- **[P1] Capital One Venture X statement** — confirm due date + autopay.
  Action: log in, verify autopay is on. ~10min. 4pm.

## ⚡ QUICK — under 5 minutes each
- [P2] Bulk archive 13 promotional emails — Slickdeals / Macy's / Princess Polly. ~10min. 9:15pm.

Sources:
[AIGC Challenge](https://mail.google.com/mail/u/0/#inbox/THREAD_ID)
[Amazon delivery](https://mail.google.com/mail/u/0/#inbox/THREAD_ID)

Tools and cost

| Item | Detail | |---|---| | Stack | Claude.ai (Pro plan) + Gmail Connector | | Monthly cost | $20 (Claude Pro) — no API key, no extra subscriptions | | Setup time | ~20 minutes one-time | | Daily trigger | Manual, ~30 seconds on your phone | | Code required | None |

If you cancel Claude Pro the agent simply stops — nothing to clean up.

Common failure modes

These are the real ones from the first two weeks of running it. Fix the prompt, not the output — every prompt edit makes tomorrow's triage permanently better.

  • A "thanks" reply lands in P0. Add a rule: replies that just acknowledge something are never P0.
  • A buried meeting confirmation gets missed. Add a rule: for any thread with more than 5 messages, always call get_thread to read the full thread before classifying.
  • Amazon/delivery notifications over-flag. Add specific senders to a "summary-only" list — return one count, not one item per message.
  • A marketing "Final notice" misread as urgent. Add a rule: senders in your skip-list cannot be P0, no matter the subject.
  • Days 1-7 are noisy. Expect ~90% accuracy only after a week of prompt tuning. Don't judge it before then.

FAQ

Where do I paste this prompt? Into Claude's Project system prompt, or as a Custom Instruction. Open Claude.ai → Settings → Custom Instructions, or create a new Project and put it in the project's system prompt slot.

Do I need a Claude API key? No. This prompt runs entirely inside Claude.ai (Pro plan) — no API, no code. The Gmail Connector is a Pro feature you authorize once.

Will the agent send or delete email? No. The Constraints section explicitly forbids it. The agent only reads, classifies, and recommends — the actions stay yours.

Can I use ChatGPT or Gemini instead? The prompt structure transfers, but the Gmail Connector is Claude-specific. ChatGPT can read Gmail via a custom GPT + Zapier, but you lose the no-code, no-API-key path.

How long until it's accurate? About 5-7 days of prompt tuning. The first week, expect to add one or two rules per morning. After that it stabilizes at roughly 90% accurate priority calls.

What if my inbox has fewer than three emails on a day? The prompt explicitly tells the agent to say "inbox is quiet" rather than invent tasks. Quiet mornings are okay.

Related Workflows

One real AI agent a week, straight to your inbox. Free, no upsell.