xiaohongshu-lowtop

Warn

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The script scripts/fetch_explosive_articles.py contains logic to scan sensitive shell configuration files, such as ~/.bashrc, ~/.zshrc, ~/.bash_profile, ~/.profile, and Windows PowerShell profiles, to automatically locate the REDFOX_API_KEY. Accessing these files can expose other sensitive secrets and credentials stored on the user's system.
  • [DATA_EXFILTRATION]: Network requests made to the redfox.hk API in scripts/fetch_explosive_articles.py use native Python socket and SSL modules with certificate verification disabled (ssl.CERT_NONE). This insecure connection strategy risks exposing the API key and fetched data to interception.
  • [EXTERNAL_DOWNLOADS]: The assets/preview-template.html file downloads the html2pdf.js library from external CDNs, including jsDelivr, Cloudflare, and unpkg. This is used for the legitimate purpose of generating PDF exports from the generated results.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from Xiaohongshu (note titles and descriptions) and displays it without sanitization.
  • Ingestion points: Data is retrieved from the redfox.hk API in scripts/fetch_explosive_articles.py.
  • Boundary markers: Absent; external content is placed directly into the markdown output without delimiters or warnings to ignore embedded instructions.
  • Capability inventory: The agent can execute Python scripts, write files, and provide downloadable attachments.
  • Sanitization: Only basic whitespace cleaning is performed for layout purposes; no security-focused filtering is present.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 12, 2026, 07:51 PM
Security Audit — agent-trust-hub — xiaohongshu-lowtop