Brian Knows — agentic threat model

Not certain from the listing — likely relies on third-party commercial LLMs for natural language processing. Primary threats include prompt injection that could manipulate the agent into generating malicious transaction payloads or misinterpreting user intent.

Not certain from the listing — retrieves on-chain data and searches resources to inform transactions. Vulnerable to data poisoning if it ingests manipulated on-chain state, malicious smart contract ABIs, or poisoned token metadata.

Translates natural language intents into actionable Web3 transactions (swaps, deployments). High risk of tool misuse or insecure tool integration if the translation logic fails to validate destination addresses, slippage parameters, or contract bytecode before presenting them to the user.

Not certain from the listing — hosted as a closed-source SaaS platform. Key threats include the exposure of RPC node API keys, lack of transaction simulation sandboxes, and potential compromise of the hosting infrastructure leading to transaction hijacking.

Not certain from the listing — no mention of transaction guardrails, simulation tools, or real-time monitoring. The lack of visible transaction dry-runs or anomaly detection increases the risk of executing unintended financial operations.

Closed-source, freemium model with no mentioned security audits, SOC2 compliance, or formal verification of the smart contracts it interacts with or deploys. This lack of visible compliance controls increases systemic risk.

Not certain from the listing — primarily acts as a direct user-to-blockchain interface, but interacting with external DeFi protocols introduces ecosystem risks such as cascading failures, smart contract exploits, or malicious liquidity pools.