webauthn-prf-wallet

Pass

Audited by Gen Agent Trust Hub on May 6, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides a robust security architecture for passkey-derived wallets. It correctly emphasizes the use of same-origin iframe isolation (via the Postmate library) to protect private keys from potential XSS or supply-chain attacks in the parent application context. The guidance includes critical security measures such as strict Content Security Policies (CSP), the use of 'allow' attributes for WebAuthn in iframes, and deterministic key derivation using the Web Crypto API.
  • [INDIRECT_PROMPT_INJECTION]: The skill describes a pattern where an isolated wallet context ingests commands (e.g., signing requests) from a potentially less-secure parent window via a cross-frame RPC mechanism.
  • Ingestion points: Data enters the isolated context through a Postmate RPC envelope in 'assets/WalletIframeSketch.ts'.
  • Boundary markers: The same-origin policy between the parent and the iframe acts as a hard boundary. The skill instructions further advise requiring a fresh WebAuthn ceremony (biometric check) for every operation to prevent automated abuse.
  • Capability inventory: The isolated context has access to a derived private key and the ability to sign Ethereum messages/transactions.
  • Sanitization: The implementation uses structured JSON parsing for RPC messages and relies on the platform-level WebAuthn user verification as the primary control against unauthorized signing.
Audit Metadata
Risk Level
SAFE
Analyzed
May 6, 2026, 10:29 PM
Security Audit — agent-trust-hub — webauthn-prf-wallet