Finance Query

Instructions

Answer the user's financial question using the unified SQLite ledger.

Data source

Primary: data/finance.db (SQLite — unified ledger of all transactions from Plaid + CSVs)

Query with: sqlite3 data/finance.db "SELECT ..."

Fallback (if DB missing): data/latest/*.json

Schema

transactions (
  id TEXT PRIMARY KEY,
  date TEXT,           -- ISO format YYYY-MM-DD
  description TEXT,
  amount REAL,         -- POSITIVE = debit (money spent), NEGATIVE = credit (income/refund)
  merchant TEXT,
  category TEXT,       -- Plaid category (may be empty for CSV-sourced)
  account_name TEXT,
  account_mask TEXT,   -- last 4 digits
  account_type TEXT,   -- depository, credit, loan, investment
  institution TEXT,
  source TEXT,         -- 'plaid' or 'csv'
  source_file TEXT
);

accounts (
  account_id TEXT PRIMARY KEY,
  name TEXT,
  type TEXT,
  subtype TEXT,
  mask TEXT,
  institution_id TEXT,
  balance_current REAL,
  balance_available REAL,
  balance_limit REAL,
  last_updated TEXT
);

Steps

1. Check if data/finance.db exists. If not, suggest running python3 scripts/ingest.py first.
2. Query the database using sqlite3 to answer the user's question.
3. Present results clearly with specific numbers, dates, and merchant names.

Query patterns

# Date range spending
sqlite3 data/finance.db "SELECT description, amount, date FROM transactions WHERE date >= '2026-06-01' AND amount > 0 ORDER BY amount DESC LIMIT 20"

# Spending by merchant
sqlite3 data/finance.db "SELECT description, SUM(amount), COUNT(*) FROM transactions WHERE amount > 0 GROUP BY description ORDER BY SUM(amount) DESC LIMIT 20"

# Monthly totals
sqlite3 data/finance.db "SELECT substr(date,1,7) as month, SUM(CASE WHEN amount > 0 THEN amount ELSE 0 END) as spent, SUM(CASE WHEN amount  100 ORDER BY date DESC LIMIT 20"

Conventions

• amount > 0 = money spent (debit)
• amount < 0 = money received (credit/income)
• Dates are ISO format (YYYY-MM-DD)
• Always clarify the time period in your response
• Round dollar amounts to 2 decimal places
• Use -header -column for readable output: sqlite3 -header -column data/finance.db "..."

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: weklund
• Source: weklund/fiduciary
• License: MIT

Install and usage instructions live in the source repository linked above.