🔍Intermediate⏱ 60 min

Build a Scam Detection Agent

You will explore the ML scam-detection pipeline (scikit-learn / LightGBM), understand the feature engineering for token addresses, deploy the model endpoint, then expose it as a priced agent skill through the x402 server.

mlscam-detectionpythonagent-skill

Prerequisites:Your First Agent on Base

🏆

Ability to build and deploy ML-powered agent skills

Step 1: Understand the feature pipeline

The scam detector extracts on-chain features like token age, holder concentration, liquidity ratio, and contract verification status. Review `ml/features/` to understand the feature set.

Step 1 — python

# ml/features/token_features.py (excerpt)
def extract_features(token_address: str) -> dict:
    return {
        'holder_count': get_holder_count(token_address),
        'top_holder_pct': get_top_holder_percentage(token_address),
        'liquidity_usd': get_liquidity(token_address),
        'contract_verified': is_contract_verified(token_address),
        'age_days': get_token_age_days(token_address),
        'transfer_count': get_transfer_count(token_address),
    }

✓

Checkpoint

Able to extract features for a known token address

1 / 3