The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes python3 -m py_compile to perform a syntax check on the generated wxbot.py file to ensure the code is valid before presentation to the user.
[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to install the wxautox4 third-party library and the python-dotenv package via pip. It also references external services for library activation and registration at https://dusapi.com and https://wxauto.org.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it is designed to ingest and act upon untrusted message data from WeChat.
Ingestion points: Untrusted data enters the agent context through WeChat message objects, specifically msg.content, within the on_message and on_new_message callbacks in wxbot.py.
Boundary markers: The provided templates do not include boundary markers, delimiters, or system instructions to ignore embedded commands within the processed messages.
Capability inventory: The generated scripts have the capability to perform various actions on the WeChat account, including SendMsg, PublishMoment, Comment, and AtAll within wxbot.py.
Sanitization: No sanitization, escaping, or validation of the external message content is implemented in the templates before the data is used to trigger responses or actions.

wxauto-dev