auto-tag

Auto-Tag Gmail Emails

Analyze emails and apply label recommendations with user confirmation.

Arguments

$ARGUMENTS — optional Gmail search query. If empty, use recent INBOX messages.

Step 1: Fetch emails and labels

1. Use Gmail MCP gmail_search_messages to fetch emails (use query from $ARGUMENTS or in:inbox, limit 10-20)
2. Use Gmail MCP gmail_list_labels to get available user labels
3. For each email, use gmail_read_message to get full content and current labels

Step 2: Classify with tagging context

For each email, classify into one of 5 categories:

• 🟢 PERSONAL / ЛИЧНОЕ — addressed personally, requires response
• 🔵 USEFUL NEWSLETTER / ПОЛЕЗНАЯ РАССЫЛКА — consciously subscribed content
• 🟡 IMPORTANT NOTIFICATION / ВАЖНОЕ УВЕДОМЛЕНИЕ — transactional, worth reading
• 🟠 NOISE / ШУМ — repetitive alerts, noise
• 🔴 SPAM / МУСОР — marketing, promotions, spam

For each email provide:

1. Category (with emoji)
2. Brief summary
3. Recommendation
4. Suggested labels from the available user label list

Step 3: Generate tagging proposal

Create a summary table of proposed changes:

• For each email: abbreviated subject, current labels, proposed new labels (by name)
• Only propose labels that are NOT already on the email

Step 4: Confirm with user

Ask the user to confirm: "Apply these tags? (y/n)"

Do NOT proceed without explicit user confirmation.

Step 5: Apply tags (only after confirmation)

After user confirms, construct JSON and pipe to the apply script:

echo '{"actions": [{"message_id": "MSG_ID", "add": ["Label_ID"]}]}' | python3 ~/.claude/scripts/gmail-analyzer/modify_labels.py

Report results: how many emails were tagged successfully.

If user declines, report "Cancelled." and do not apply.

Respond in the user's language.