Threat Model
Autonomous agents execute actions on behalf of users. When financial transactions are involved, the attack surface expands significantly. x402Shield protects against these vectors:
Enforcement Gate
Indirect Prompt Injection
An attacker injects malicious instructions into untrusted context (e.g. summarizing a webpage or reading an email) that tricks the agent into initiating a payment or extracting funds.
Malicious MCP/Tool Responses
A compromised or malicious tool returns crafted output designed to manipulate the agent's next action, altering payment amounts or destinations.
Recipient Substitution
The agent is convinced to send a valid payment to an attacker-controlled wallet instead of the intended recipient.
Price Manipulation
The agent is tricked into paying significantly more than the market rate or approved budget for a resource.
Payment Proof Replay
A valid x402 payment proof is intercepted and reused by an attacker to access paid resources without authorization.
Concurrent Reuse
A single payment proof is submitted simultaneously to multiple nodes to bypass usage limits.
Request–Resource Mismatch
A valid payment proof for one endpoint or resource is used to authorize access to a different, potentially more expensive endpoint.
Budget Abuse
An agent drains a wallet through repeated, high-frequency microtransactions that individually appear valid but exceed session or global limits.
Sensitive Context Leakage
The agent unwittingly includes private keys, PII, or sensitive context in the body of an outbound request or transaction memo.