linkfox-zhihuiya-abstract-data-translated
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFEPROMPT_INJECTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes patent abstracts and titles retrieved from the external PatSnap database via an API call. This content is untrusted and represents an indirect prompt injection surface where malicious instructions could be embedded in patent data to target the agent's logic.
- Ingestion points: Patent titles and abstracts fetched through the
/zhihuiya/abstractDataTranslatedendpoint inscripts/zhihuiya_abstract_translated.py. - Boundary markers: The skill does not employ explicit delimiters or warnings to ignore instructions within the retrieved patent data when displaying results.
- Capability inventory: The skill scripts have the ability to write JSON logs to the local workspace and communicate with vendor APIs.
- Sanitization: No specific sanitization or filtering logic for the retrieved patent text is implemented in the scripts before display.
- [DATA_EXPOSURE]: The
references/onboarding.mdfile provides instructions for users to store theirLINKFOX_AGENT_API_KEYin plain text inside shell profile files such as~/.zshrc,~/.bashrc, and~/.profile. This practice stores sensitive credentials in persistent system configuration files, which increases the risk of credential exposure to other local processes or during session auditing. - [EXTERNAL_DOWNLOADS]: The
scripts/onboarding.pyutility script includes logic to detect missing third-party dependencies (requests,qrcode,pillow) and provides the user with shell commands to install them via the system package manager.
Audit Metadata