The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill's main execution script automatically reads sensitive shell profile files, such as ~/.zshrc, ~/.bashrc, and Windows PowerShell profiles, to extract API keys. These files frequently contain other sensitive environment variables and credentials, and accessing them constitutes an unsafe data exposure practice. Evidence: get_redfox_api_key function in scripts/fetch_explosive_articles.py scans multiple system-level configuration paths.
[DATA_EXPOSURE_AND_EXFILTRATION]: The network communication code explicitly disables SSL certificate verification, creating a vulnerability to man-in-the-middle (MITM) attacks that could expose the API key or transmitted data. Evidence: fetch_ranking_data in scripts/fetch_explosive_articles.py sets context.check_hostname = False and context.verify_mode = ssl.CERT_NONE.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external social media API and interpolates it into reports, creating an indirect prompt injection surface. Evidence Chain: 1. Ingestion points: fetch_ranking_data in scripts/fetch_explosive_articles.py reads titles and descriptions from the RedFox API. 2. Boundary markers: Absent; note content is interpolated without delimiters. 3. Capability inventory: Includes network operations, file writing, and shell command execution. 4. Sanitization: Absent; only basic whitespace removal is performed on ingested text.

xiaohongshu-lowtop