metamask-agent-wallet
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill documents the MetaMask Agentic CLI ('mm') and provides helper scripts for blockchain operations. It emphasizes security best practices throughout its instructions.
- [CREDENTIALS_UNSAFE]: The skill correctly instructs users and agents never to pass sensitive data like mnemonics or passwords as inline flags. Instead, it mandates the use of environment variables (e.g., 'MM_MNEMONIC', 'MM_PASSWORD') to prevent secrets from appearing in shell history or process logs.
- [COMMAND_EXECUTION]: The skill uses 'subprocess.run' in 'scripts/x402_pay.py' to interface with the vendor's own 'mm' CLI. This execution is limited to the skill's internal logic and is used for secure cryptographic signing operations where the private key remains managed by the wallet.
- [INDIRECT_PROMPT_INJECTION]: While the skill ingests external data (such as HTTP 402 challenges and MCP tool results), it mitigates potential risks through strict 'Confirmation Requirements'. The agent is explicitly instructed to show the asset, amount, network, and recipient to the user and obtain manual approval before any authorization is signed or funds are moved.
- [DATA_EXFILTRATION]: The skill includes safety checks in its payment script, such as a 'no-redirect' handler and a protocol check that enforces HTTPS (except for local loopback testing), preventing the accidental diversion of payment authorizations to unauthorized hosts.
Audit Metadata