Email Summary Example

TODO: Perform the task manually with the agent in the chat window and ask it to turn it into a project skill in this solution.

Email Summary

EXAMPLE: Fetch today's inbox via the Gmail MCP and produce a privacy-safe digest with prioritized next actions.

Steps

EXAMPLE:

1. Determine today's date in the user's local timezone. If currentDate is provided in context, use it; otherwise use the current date.
2. Call mcp__claude_ai_Gmail__search_threads with:
   • query: after:YYYY/MM/DD before:YYYY/MM/DD in:inbox (upper bound is the day after today)
   • pageSize: 50
3. Count individual messages whose date field falls on today's date (a returned thread may include older messages — do not count those).
4. Group today's messages into categories based on sender domain and subject patterns. Pick buckets that fit the actual data, e.g.:
   • Platform / community notifications (circle.so, Discord, Slack, GitHub, etc.)
   • Customer or subscriber support
   • Calendar invites and self-forwards
   • Newsletters / promotional
   • Direct human correspondence
5. Anonymize. Never echo sender names, email addresses, or verbatim subject lines. Refer to senders generically ("a long-time subscriber", "a program participant", "a notification platform"). Describe subjects by topic, not by their literal text. Do not include any names that appear inside snippets either.
6. Pick the top 3 to focus on next using these signals, in order:
   • Direct human messages waiting on a reply from the user
   • Paying customers blocked, frustrated, or asking a question
   • Time-sensitive items tied to events today or tomorrow
   • Skip routine notifications, automated digests, and self-sent mail

Output format

EXAMPLE:

## Today's Email Summary (<N> messages)

### By category

**<Category> — <count>**
<one short paragraph, anonymized>

**<Category> — <count>**
...

### Top 3 to focus on next

1. **<Short label>** — <why it matters and what action is needed>
2. **<Short label>** — ...
3. **<Short label>** — ...

Keep the whole response tight: short paragraph per category, one-liner per priority item with a clear action.