The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script fetches article content and images from WeChat's official domains (mp.weixin.qq.com, mmbiz.qpic.cn). These are well-known services and the behavior is essential to the skill's primary purpose.
[COMMAND_EXECUTION]: No evidence of arbitrary command execution. The script uses standard libraries for its operations and does not invoke shell commands with untrusted input.
[DATA_EXFILTRATION]: No sensitive local data is accessed or transmitted. The network activity is limited to fetching the requested article content.
[PROMPT_INJECTION]: The skill processes external HTML content, which presents a surface for indirect prompt injection. However, the risk is localized as the script does not use high-privilege capabilities (like eval or subprocess) to process this data.
Ingestion points: scripts/fetch_wechat_article.py fetches external HTML content via the requests library.
Boundary markers: No specific boundary markers are used when presenting the content to the agent context.
Capability inventory: The skill is capable of performing local file system writes (Path.mkdir, Path.write_text, Path.write_bytes) and network GET requests.
Sanitization: The safe_filename function uses regular expressions to strip dangerous characters (e.g., \ / : * ? " < > |) from article titles and account names before using them in file paths, mitigating directory traversal risks.

wechat-article-fetcher