The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the iron-session package from the official NPM registry to handle encrypted cookie-based sessions. This is a standard dependency for the stated functionality.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external Ethereum wallets.
Ingestion points: The /api/siwe/verify API route in packages/nextjs/app/api/siwe/verify/route.ts accepts message and signature strings from the request body.
Boundary markers: The implementation lacks explicit boundary markers or instructions to the agent to ignore potentially malicious content embedded within user-controlled fields of the SIWE message, such as the statement or uri.
Capability inventory: The skill uses the ingested data for signature verification. Although it does not directly perform dangerous system operations using this data, the resulting authenticated session state determines the user's identity and permissions for subsequent agent actions.
Sanitization: The implementation uses the viem/siwe library to parse and verify the message, which provides structural and cryptographic validation but does not sanitize the natural language content for potential prompt injection attacks.

siwe