The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run and subprocess.Popen extensively to orchestrate workflows between its internal scripts and the external wcx CLI tool. All calls use argument lists rather than shell strings, mitigating command injection risks.
[EXTERNAL_DOWNLOADS]: The skill identifies a dependency on the wcx library hosted on GitHub (lovstudio/wcx). It follows a safe pattern by pinning the dependency to a specific git commit and instructing the user to install it manually rather than performing silent automatic installations.
[DATA_EXFILTRATION]: The skill includes a fallback mechanism that sends WeChat article URLs to the Metaso API (https://metaso.cn) for content extraction. This behavior is documented and guarded by a mandatory --allow-metaso flag, ensuring it only occurs with explicit user authorization.
[SAFE_PRACTICES]: Several defensive measures are implemented:
URL Validation: The normalize_wechat_url function strictly enforces https://mp.weixin.qq.com/s... patterns and strips tracking parameters to prevent malicious redirection or data leakage.
Permission Hardening: The secure_directory utility applies 0o700 permissions (read/write/execute for owner only) to data and profile directories.
Input Sanitization: Filenames are sanitized using safe_component to prevent path traversal and shell meta-character issues.

wxmp-article-harvester