The Agent Skills Directory

[SAFE]: The skill uses a legitimate mechanism for managing sensitive credentials by loading them from environment variables via .env files in a vendor-specific directory (~/.wmyskills/.env).
[COMMAND_EXECUTION]: The script scripts/freshrss.py uses subprocess.run to execute the Windows command chcp 65001. This is a benign operation intended solely to set the terminal encoding to UTF-8 for proper character display; it uses a hardcoded string and does not accept user input.
[DATA_EXPOSURE]: The skill communicates with a user-defined FreshRSS instance via the Google Reader API. This is the primary function of the skill. While it processes external RSS feed data which could contain malicious text (indirect prompt injection), the skill implements HTML sanitization (_strip_html) and lacks any dangerous capabilities (such as file writing or arbitrary code execution) that could be exploited by such content.

freshrss