use-user-controlled-wallets

Pass

Audited by Gen Agent Trust Hub on Apr 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill contains a dedicated security section with non-negotiable rules for the agent, including the requirement for human confirmation before executing any mainnet transactions and the use of environment variables for secrets.- [EXTERNAL_DOWNLOADS]: The skill installs and uses official vendor packages from the @circle-fin namespace and well-known development utilities. These are recognized as legitimate resources provided by the authoring organization.- [DATA_EXFILTRATION]: While the skill handles sensitive session tokens (userToken, encryptionKey), it explicitly warns against insecure storage (localStorage) in production and provides instructions for using httpOnly cookies to prevent cross-site scripting (XSS) risks.- [PROMPT_INJECTION]: A surface for indirect prompt injection exists where the skill processes user-supplied emails and wallet addresses. However, this is mitigated by instructions requiring strict input validation and the use of a hosted, out-of-band UI for all cryptographic signing operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 24, 2026, 09:02 PM