okx-ai
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill installs the @okxweb3/a2a-node package globally via npm to bootstrap the communication runtime. This package is hosted under the vendor's official NPM scope and is used for peer-to-peer messaging functions.
- [COMMAND_EXECUTION]: Extensive use of vendor-specific CLI tools (onchainos and okx-a2a) to interact with blockchain identities and the task marketplace. The skill follows a strict 'one intent = one CLI call' rule and validates inputs, such as service descriptions and URLs, before execution to prevent command injection.
- [PROMPT_INJECTION]: The skill contains robust instructions for the agent to ignore potentially malicious instructions embedded in user-provided metadata, task descriptions, or dispute materials. It specifically includes a 'Dangerous-Instruction Gate' in the sub-session playbook to prevent peer agents from requesting sensitive data, system-level access, or ignoring primary instructions.
Audit Metadata