receipt-email

/smartspender:receipt email

Trigger

• /smartspender:receipt email
• /smartspender:receipt email 30 (scan last 30 days)
• "Scan mine emails for kvitteringer"
• "Hent fakturaer fra email"

Arguments

Argument	Required	Values	Default
action	yes	email	-
days	no	positive integer	last_email_scan or 90

Prerequisites

• Gmail MCP server configured and accessible
• At least one bank sync completed (transactions.csv exists) for matching to work

Workflow

1. Check Gmail MCP availability. If not available, respond: "Gmail MCP er ikke konfigureret. Tilfoej Gmail MCP-serveren til dit Cowork-miljoe for at kunne scanne emails."
2. Determine scan date range:
   • If days argument provided: scan from (today - days) to today
   • If last_email_scan exists in settings.csv: scan from that timestamp to today
   • Otherwise: scan last 90 days
3. Load email scanning rules: skills/email-receipt-scanning/SKILL.md
4. Search Gmail using the primary search query with the calculated date range
5. If primary query returns fewer than 3 results, also run supplementary queries
6. Filter results using the include/exclude heuristics from email-receipt-scanning skill
7. If no emails found, respond: "Ingen kvitterings- eller fakturaemails fundet i den valgte periode."
8. Present found emails to user:

   Jeg fandt {n} emails med kvitteringer eller fakturaer:
   
   1. {date} — {sender} — "{subject}" {attachment_info}
   2. {date} — {sender} — "{subject}" {attachment_info}
   3. {date} — {sender} — "{subject}" {attachment_info}
   ...
   
   Vil du importere alle, vaelge enkelte, eller annullere? (alle/vaelg/annuller)
   

9. [USER ACTION]: User chooses "alle", "vaelg" (then picks numbers), or "annuller"
10. If user cancels, respond: "Email-scanning annulleret." and stop.
11. Load document-parsing skill: skills/document-parsing/SKILL.md
12. Load extraction and vendor detection rules from skills/document-parsing/SKILL.md
13. Load schema: skills/data-schemas/SKILL.md
14. Load matching rules: skills/transaction-matching/SKILL.md
15. For each selected email, process in sequence: a. Detect content type (PDF attachment, inline HTML, or both — prefer PDF) b. If PDF: download attachment c. Detect vendor from email sender domain (per email-receipt-scanning skill sender mapping) or fall through to invoice-parsing skill vendor detection d. Check for vendor-specific parser: invoice-knowledge/{vendor-id}/PARSER.md e. Extract receipt metadata and line items using vendor parser or general rules f. Search transactions.csv for matching transactions (amount +-1%, date +-1 day) g. Score candidates per transaction-matching confidence rules h. If exactly 1 match with confidence >= 0.8: auto-link i. If ambiguous: mark as ambiguous (batch mode — don't interrupt for each email) j. Check for duplicate in receipts.csv (date + merchant + total_amount) k. If duplicate: skip this email and note as "allerede registreret" l. Generate receipt_id and item_ids m. If PDF: save to receipts/{receipt_id}.pdf, set file_reference to this path n. If inline HTML only: set file_reference to email:{message_id} o. Set source: email on the receipt row p. Append receipt row to receipts.csv q. Determine the monthly file from the receipt date: receipt-items-{YYYY-MM}.csv r. If the monthly file does not exist, create it with the header row s. Append item rows to receipt-items-{YYYY-MM}.csv t. Log to action-log.csv
16. Update last_email_scan in settings.csv to current datetime. If settings.csv does not exist, create it with key,value header and the last_email_scan row.
17. Output batch summary