The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md instructions explicitly direct the agent to automatically modify the user's shell configuration files (~/.zshrc, ~/.bashrc) or use PowerShell commands to set permanent environment variables (SetEnvironmentVariable) if the REDFOX_API_KEY is missing. This constitutes a persistence mechanism and automated modification of system configuration.
[DATA_EXFILTRATION]: The check_sensitive_words.py script transmits user-provided content to a remote API (https://redfox.hk/story/api/cozeSkill/sensitiveWordSearch) for analysis. While the skill identifies this as its core function, the transmission of potentially sensitive text to a third-party service represents a data exposure risk.
[EXTERNAL_DOWNLOADS]: The extract_text.py script uses the requests library to fetch and scrape content from arbitrary URLs provided by the user. This involves processing untrusted external data which could lead to indirect prompt injection or exposure to malicious web content.
[CREDENTIALS_UNSAFE]: The check_sensitive_words.py script contains logic to search for and read the REDFOX_API_KEY directly from shell configuration files (.bashrc, .zshrc, etc.) if not found in the environment variables, which is an aggressive credential harvesting pattern.

multi-wordcheck