byted-kickart-video-analyzer
Fail
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent to prompt the user to provide their 'ACCESS_KEY_ID' and 'SECRET_ACCESS_KEY' in the chat if they are not configured. It then provides instructions to execute an 'export' command to load these secrets into the environment, which exposes sensitive credentials in the session history.\n- [REMOTE_CODE_EXECUTION]: During the '技能版本校验' (Skill Version Check) step, the skill fetches data from a remote API (icp.volcengineapi.com) which includes an 'install_command' field. The instructions direct the agent to execute this arbitrary command upon user confirmation, creating a vector for remote code execution.\n- [COMMAND_EXECUTION]: The skill makes extensive use of shell command execution, including using 'curl' to download files from user-provided URLs and running local Python scripts ('plan.py', 'upload.py', 'analyze.py', 'upgrade.py') with various shell arguments.
Recommendations
- AI detected serious security threats
Audit Metadata