The Agent Skills Directory

[DYNAMIC_EXECUTION]: The scripts/article_tools.py script employs dynamic module loading by modifying sys.path to include a local toolkit directory and then importing WeChatConverter and load_theme. This runtime loading from computed paths is a significant dynamic execution pattern.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external inputs like news links, PDFs, and user-provided materials to generate content, creating a surface for injection.
Ingestion points: User-provided materials, external news URLs, and PDF documents ingested during the article writing process.
Boundary markers: The skill lacks explicit structural delimiters or safety warnings for the interpolation of untrusted external data into the agent's prompts.
Capability inventory: The associated script can read and write local files (manifests, previews) and perform network requests to the WeChat API.
Sanitization: There is no evidence of filtering or escaping external content before it is used to influence the generative output.
[COMMAND_EXECUTION]: The skill uses a bundled Python script (scripts/article_tools.py) to perform its core logic, including extensive file system operations and managing WeChat draft synchronization. While functional, this represents high-privilege tool usage for managing user data.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to api.weixin.qq.com for authentication and draft management. These operations target a well-known service required for the skill's primary functionality and are documented as part of the operational workflow.

zg-wechat-article