Agentic, Context-Aware Risk Intelligence in the Internet of Value

Risk across blockchains is no longer a property of any single chain. It comes from a combination of route, sentiment, liquidity and the policy a system commits to.

The paper proposes a risk system built from five engines: prediction, decentralised verification, sentiment fusion, an agent that acts under written rules, and a scenario engine that turns forecasts into pre-committed action programs.

It supports the design with two case studies: a 27-hour liquidity stress-response experiment on Solana, and a 168-hour calibration of the prediction router, reported with explicit class-imbalance caveats.