The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The scripts check_sensitive_words.py, fetch_xhs_trends.py, and fetch_explosive_covers.py are designed to search for the REDFOX_API_KEY by reading sensitive shell configuration files, including ~/.zshrc, ~/.bashrc, ~/.bash_profile, ~/.profile, and ~/.zprofile. Accessing these files is a security concern as they frequently contain various sensitive credentials and environment variables.
[COMMAND_EXECUTION]: The skill executes PowerShell commands on Windows systems via subprocess.run within fetch_xhs_trends.py and fetch_explosive_covers.py to retrieve environment variables.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the playwright library and the Chromium browser binary to support web scraping functionality for prohibited word detection.
[DATA_EXFILTRATION]: User-provided text and metadata regarding content rewrites are sent to vendor-controlled endpoints at redfox.hk for analysis and tracking.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources for prohibited word detection, creating a potential attack surface. \n
Ingestion points: check_sensitive_words.py (via file and URL inputs) and the agent's image text extraction capability. \n
Boundary markers: None identified in the processing scripts. \n
Capability inventory: Subprocess execution, network requests to vendor APIs, and local file writing for optimized text. \n
Sanitization: No specific sanitization or filtering of external content is performed before processing or transmission to the API.

xiaohongshu-creator