xiaohongshu-darkhorse
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill attempts to locate the REDFOX_API_KEY by scanning local shell configuration files such as ~/.zshrc, ~/.bashrc, ~/.profile, and PowerShell profiles. This is a documented convenience feature intended to assist users in providing the required API credentials.\n- [EXTERNAL_DOWNLOADS]: The skill makes authenticated POST requests to the vendor's official API at https://redfox.hk to retrieve engagement data, follower counts, and post details. This network activity is consistent with the core functionality of the skill.\n- [COMMAND_EXECUTION]: The data fetching script fetch_darkhorse_articles.py utilizes raw socket and ssl modules for network communication and explicitly disables SSL certificate and hostname verification. This practice represents a security weakness but is used to fulfill the skill's design goal of operating without external Python dependencies like 'requests'.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of Xiaohongshu post titles and descriptions to generate markdown tables and content analysis. \n
- Ingestion points: Data is retrieved from the redfox.hk API in fetch_darkhorse_articles.py.\n
- Boundary markers: Data is encapsulated within Markdown tables and HTML templates.\n
- Capability inventory: The skill performs local file operations (caching) and network requests to the vendor's API.\n
- Sanitization: Implements basic text cleaning to remove white space and line breaks but lacks comprehensive character escaping.
Audit Metadata