The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/generate_and_send.py to invoke a PowerShell script for desktop automation. This is the primary function of the skill. The implementation is secure as it converts user-supplied strings into arrays of Unicode integers (using ord(c)) before passing them to the shell, which prevents command injection attacks.
[EXTERNAL_DOWNLOADS]: The README.md file includes a tracking pixel from visitor-badge.laobi.icu for repository visitor statistics. While automated scanners may flag this domain due to its tracking nature or past reputation, it is a common practice in open-source documentation and does not impact the operational security of the skill's code.
[INDIRECT_PROMPT_INJECTION]: The skill has a data ingestion surface in scripts/generate_and_send.py where it accepts contact names and message content.
Ingestion points: sys.argv inputs in scripts/generate_and_send.py passed to scripts/send_message.ps1.
Boundary markers: None identified in the prompt templates.
Capability inventory: Shell execution (subprocess.run), UI automation (SendKeys), and mouse control (mouse_event).
Sanitization: The skill converts all input text into comma-separated Unicode code points (integers) before execution, ensuring the data is not interpreted as shell commands or control characters by the PowerShell interpreter.

wechat-sender