analyze

/smartspender:analyze

Trigger

• /smartspender:analyze
• "Analyser mine transaktioner"
• "Kategoriser mit forbrug"
• "Analyze my spending"

Arguments

None.

Prerequisites

• Transactions synced (transactions.csv has data)

Workflow

1. Load user memory skill:

   • Load skills/user-memory/SKILL.md for learning detection
   • Read learnings/categorization.md for previously learned category corrections
   • Read learnings/subscriptions.md for subscription confirmations/denials
   • Read learnings/merchants.md for merchant aliases

2. Read all transactions from transactions.csv

3. Read existing categorized transactions from categorized.csv (by tx_id)

4. Read merchant-overrides.csv (if it exists) — these are learned categorization rules from previous user corrections

5. Identify uncategorized transactions: tx_ids present in transactions.csv but not in categorized.csv

6. If no uncategorized transactions found, skip to step 8 (still check for new overrides to learn)

7. For each uncategorized transaction, apply categorization using skills/categorization/SKILL.md: a. Normalize the raw_text for pattern matching b. Check merchant-overrides.csv first — if raw_text (normalized) matches a raw_pattern, use that override's merchant/category/subcategory (confidence 1.0) c. If no override match, check against the merchant pattern database (exact match -> confidence 1.0) d. If no exact match, try partial pattern match (confidence 0.8) e. If no pattern match, use intelligent classification from transaction context (confidence 0.5-0.7) f. If still unmatched, assign to "Andet" category (confidence 0.0) g. Detect transaction type from description prefix (Dankort-kob, PBS, Overforsel, etc.) h. For known subscription merchants, set is_recurring to TRUE

8. Append categorized rows to categorized.csv (create file with header row if it doesn't exist)

9. Learn from manual corrections: scan categorized.csv for rows where manual_override is TRUE. For each, look up the original raw_text in transactions.csv. If no corresponding raw_pattern exists in merchant-overrides.csv yet, append a new override row with the corrected merchant/category/subcategory

10. Run subscription detection using skills/subscription-detection/SKILL.md: a. Group categorized transactions by merchant b. For each merchant with >=3 occurrences, check the 5 subscription criteria c. Calculate frequency and annual cost for detected subscriptions

11. Read existing subscriptions from subscriptions.csv

12. For newly detected subscriptions not already in the file, append rows

13. For existing subscriptions, update last_seen date and amount if changed

14. Calculate monthly aggregations and write to monthly-summary.csv: a. Group by month (YYYY-MM) and category b. Compute total, transaction_count, avg_transaction c. Compute vs_prev_month and vs_prev_month_pct where previous month data exists

15. Append the analysis event to action-log.csv:

    • action_type: analyze
    • target: all
    • status: completed
    • details: "{N} transactions categorized, {M} subscriptions detected"