The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides templates to build bots that ingest untrusted data from WeChat messages and moments. These bots possess capabilities to send messages and download files, creating an attack surface where malicious input could trigger unintended actions.
Ingestion points: In SKILL.md templates, data is ingested through wx.GetNextNewMessage, wx.GetAllMessage, and wx.Moments().GetMoments().
Boundary markers: Absent. The generated code templates do not include delimiters or instructions to distinguish between user data and bot instructions.
Capability inventory: In SKILL.md templates, the bots have access to wx.SendMsg, wx.PublishMoment, and msg.download() functions.
Sanitization: Absent. No input validation, escaping, or filtering of message content is implemented in the provided templates before the data is processed or used in responses.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the wxautox4, python-dotenv, and schedule packages from public registries. It also directs users to external domains wxauto.org and dusapi.com for account registration, license purchase, and API services.
[COMMAND_EXECUTION]: The skill instructions involve executing shell commands on the user's system for environment setup and script validation, specifically pip install, wxautox4 -a for licensing activation, and python3 -m py_compile for syntax checking.

wxauto-dev