Gmail Digest Skill

Process Gmail inbox and create actionable digest with priority classification.

Usage

code

/digest              # Process new emails since last run
/digest --initial    # Process last 60 days (first-time setup)
/digest --days=7     # Process last 7 days

Processing Pipeline

Execute these steps in order:

Step 1: Determine Time Range

code

If --initial flag or no previous digest exists:
  - Query: "newer_than:60d"
Else if --days=N specified:
  - Query: "newer_than:Nd"
Else:
  - Check last digest date from Supabase
  - Query emails since that date

Step 2: Fetch Emails from Gmail MCP

Use mcp__gmail__search_emails with appropriate query:

•Get up to 100 emails per batch
•Store: id, subject, from, date, snippet

Step 3: Filter Already Processed

Query Supabase processed_emails table:

sql

SELECT gmail_id FROM processed_emails WHERE gmail_id IN (...)

Remove any emails already in the database.

Step 4: Load and Apply Blacklist

Query Supabase blacklist table:

sql

SELECT email_pattern FROM blacklist

Filter out emails where sender matches ANY blacklist pattern (case-insensitive substring match).

Step 5: Classify Remaining Emails

Apply classification rules IN ORDER (first match wins):

Priority	Category	Rule
CRITICAL	rav	Sender contains `@vd.zh.ch`, `@zh.ch`, or `jobroom`
CRITICAL	billing	Subject contains: payment fail, overdue, invoice, past due, billing, subscription
HIGH	action_required	Subject contains: action required, urgent, reminder, deadline, expir
MEDIUM	personal	Sender is human (not matching automated patterns)
LOW	other	Everything else

Automated sender patterns (in addition to blacklist):

•Contains noreply, no-reply, no_reply
•Contains notifications@, alerts@, updates@
•Domain starts with mail., email., news.

Step 6: Read Full Email Content for Important Emails

For CRITICAL and HIGH priority emails:

•Use mcp__gmail__read_email to get full content
•Extract key information for draft generation

Step 7: Generate Draft Responses

For emails with needs_response = true (RAV, personal, action_required):

RAV emails (formal German/English):

code

Sehr geehrte Damen und Herren,

Vielen Dank für Ihre Nachricht vom [DATE].
[CONTEXTUAL RESPONSE BASED ON EMAIL CONTENT]

Mit freundlichen Grüssen,
Simon Yang

Personal emails (friendly):

code

Hi [FIRST_NAME],

Thanks for reaching out.
[CONTEXTUAL RESPONSE]

Best,
Simon

Action required (acknowledge):

code

Hi,

I've received this and will [ACTION] by [REASONABLE_DEADLINE].

Best,
Simon

Save drafts to Supabase draft_responses table.

Step 8: Build Markdown Digest

Create digest with this structure:

markdown

# Email Digest - [DATE]

**Summary**: X emails processed, Y critical, Z high priority

---

## CRITICAL (Immediate Action Required)

### RAV / Government
- **[SUBJECT]** from [SENDER] ([DATE])
  - [SNIPPET/SUMMARY]
  - Action: [WHAT NEEDS TO BE DONE]
  - Draft response prepared: [YES/NO]

### Billing / Payments
- **[SUBJECT]** from [SENDER] ([DATE])
  - [SNIPPET/SUMMARY]
  - Action: [PAYMENT NEEDED / CHECK ACCOUNT]

---

## HIGH PRIORITY

### Action Required
- **[SUBJECT]** from [SENDER] ([DATE])
  - [SNIPPET/SUMMARY]
  - Deadline: [IF MENTIONED]

---

## MEDIUM PRIORITY

### Personal Communications
- **[SUBJECT]** from [SENDER] ([DATE])
  - [SNIPPET/SUMMARY]
  - Response needed: [YES/NO]

---

## LOW PRIORITY (FYI)
- [SUBJECT] from [SENDER] - [ONE LINE SUMMARY]

---

## Draft Responses Prepared

| Email | Status | Preview |
|-------|--------|---------|
| RE: [SUBJECT] | Pending | [FIRST 50 CHARS]... |

---

## Statistics
- Total processed: X
- Critical: Y
- High: Z
- Medium: A
- Low: B
- Blacklisted/filtered: C

Step 9: Save Digest

•Save to Supabase digests table:

sql

INSERT INTO digests (date, content, email_count, critical_count, high_count)
VALUES ([TODAY], [MARKDOWN_CONTENT], [COUNT], [CRITICAL], [HIGH])
ON CONFLICT (date) DO UPDATE SET content = EXCLUDED.content, ...

•Save to filesystem:

code

docs/digests/YYYY-MM-DD.md

Step 10: Mark Emails as Processed

Insert all processed emails into Supabase:

sql

INSERT INTO processed_emails (gmail_id, thread_id, subject, sender, sender_email, received_at, priority, category, needs_response, snippet, digest_date)
VALUES ...

Step 11: Report Summary

Output summary to user:

code

Digest complete for [DATE]

Processed: X emails
- Critical: Y (RAV: a, Billing: b)
- High: Z
- Medium: A
- Low: B

Filtered: C emails (blacklisted)
Draft responses: D prepared

View digest: docs/digests/YYYY-MM-DD.md
Dashboard: http://localhost:3000

Classification Details

RAV Detection

Swiss unemployment office (Regionales Arbeitsvermittlungszentrum):

•@vd.zh.ch - Volkswirtschaftsdirektion Kanton Zürich
•@zh.ch - Canton Zurich general
•jobroom.ch - Swiss job portal

Billing Detection

Keywords in subject (case-insensitive):

•payment fail, payment failed, payment failure
•overdue, past due, past-due
•invoice, invoice #, invoice number
•billing, bill due
•subscription expir, account suspend
•action required + (payment|account|billing)

Personal Detection

NOT matching:

•Any blacklist pattern
•Automated sender patterns (noreply, notifications, etc.)
•Known marketing domains
•Bulk mail indicators

Error Handling

•If Gmail MCP fails: Report error, suggest checking connection
•If Supabase fails: Save digest to filesystem only, warn user
•If no new emails: Report "No new emails since last digest"
•If all emails blacklisted: Report count, suggest reviewing blacklist

Maintenance Commands

code

/digest --blacklist-add "pattern" "reason"   # Add to blacklist
/digest --blacklist-remove "pattern"         # Remove from blacklist
/digest --stats                              # Show processing statistics