Opus finding
Float-based USDC conversion can truncate or inflate raw units
mediumbughigh
- agenticbets/scripts/agenticbets.py:195
Multiplying a float by 1e6 then int() truncates and can be off by one due to IEEE-754. E.g., float('0.1')*1e6 = 100000.00000000001 → 100000 (fine), but float('1.1')*1e6 = 1100000.0000000001 → 1100000 (fine); however float('2.3')*1e6 = 2299999.9999999995 → 2299999, costing the user one micro-USDC and producing surprising on-chain amounts. For a financial CLI, Decimal should be used. Worse: there's no validation that amount > 0 or that the parsed float succeeded; negative or zero amounts would still encode and submit.
Recommendation
Use `decimal.Decimal(amount) * Decimal(10**USDC_DECIMALS)` and quantize; reject amount_raw < 1_000_000 (the documented minimum) before submitting any approve tx.