agent-wallet
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill manages a wallet keystore and policy configuration stored locally in
~/.chaingpt-mcp/agent-wallet/. This access is necessary for the skill's documented purpose of providing the agent with a secure, policy-bounded wallet. - [COMMAND_EXECUTION]: The skill utilizes system-level CLI tools (
securityon macOS andsecret-toolon Linux) to securely store and retrieve passphrases from the operating system's keychain, which is a recommended security practice for handling sensitive credentials. - [PROMPT_INJECTION]: The instructions explicitly address prompt injection risks by establishing a 'trust boundary' at the tool code level. It ensures that security policies are enforced by deterministic logic that remains outside the influence of the LLM's context.
- [SAFE]: The skill includes a local administrative dashboard bound to
127.0.0.1. It implements multiple security layers for this dashboard, including one-time admin tokens, CSRF protection via Origin checks, and strict schema validation for policy updates.
Audit Metadata