linkfox-ruiguan-utility-patent-detection
Fail
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill's onboarding documentation (references/onboarding.md) provides specific commands for the agent to modify the user's shell configuration files, including ~/.zshrc and ~/.bashrc, to export and persist the LINKFOX_AGENT_API_KEY environment variable. This constitutes a persistence mechanism targeting sensitive system files.
- [EXTERNAL_DOWNLOADS]: The authentication and billing management script (scripts/onboarding.py) requires the installation of third-party Python packages (requests, qrcode, pillow) that are not included in the standard library.
- [DATA_EXFILTRATION]: The skill transmits product titles and descriptions to external vendor-controlled domains (tool-gateway.linkfox.com, api.linkfox.com) to perform patent analysis and handle user authentication. While functional for the skill's purpose, it involves sending potentially proprietary product information to a remote server.
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting untrusted product information that is then processed by scripts with network and file-system capabilities.
- Ingestion points: productTitle and productDescription parameters defined in SKILL.md.
- Boundary markers: None identified; input data is interpolated into API requests without explicit delimiters or warnings to ignore embedded instructions.
- Capability inventory: The skill possesses network access capabilities (urllib, requests) and file-write capabilities (scripts/ruiguan_utility_patent_detection.py, scripts/onboarding.py).
- Sanitization: The input data is serialized via JSON for transmission, providing basic character escaping but lacking semantic validation of the content.
Recommendations
- AI detected serious security threats
Audit Metadata