xiaohongshu-weeklytop

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from an external API. \n
  • Ingestion points: Xiaohongshu post titles and descriptions are fetched via xhs_weekly_fetcher.py from the redfox.hk API. \n
  • Boundary markers: Absent; data is processed and analyzed without delimiters or warnings to the model to ignore embedded instructions. \n
  • Capability inventory: The skill executes local Python scripts, writes to the filesystem (JSON cache and HTML reports), and performs network requests. \n
  • Sanitization: Absent; no content filtering or sanitization is applied before data is used in analysis or reports. \n- [DATA_EXFILTRATION]: The xhs_weekly_fetcher.py script includes a fallback mechanism that reads sensitive shell configuration files (e.g., ~/.zshrc, ~/.bashrc, .bash_profile) to find the REDFOX_API_KEY. While documented as a convenience feature, accessing these files is a high-privilege operation. \n- [COMMAND_EXECUTION]: The skill relies on the execution of Python scripts to interact with external APIs and generate visualization reports. These scripts use the requests library to communicate with the vendor's API and perform file system writes for caching and report generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 07:51 PM
Security Audit — agent-trust-hub — xiaohongshu-weeklytop