The Agent Skills Directory

[SAFE]: The skill performs legitimate archival of WeChat articles to a local Obsidian vault as described in the documentation.
[EXTERNAL_DOWNLOADS]: The skill downloads content from api.weixin.qq.com and mp.weixin.qq.com. These are well-known domains associated with the WeChat platform and are used as expected for the skill's purpose.
[CREDENTIALS_UNSAFE]: The skill requires sensitive authentication data such as API keys and session cookies. It correctly instructs the user to manage these through a local configuration file (config.yml) in a private directory, avoiding hardcoding or accidental exposure in logs or shared files.
[DATA_EXFILTRATION]: Data access is limited to fetching articles from the WeChat platform and writing them to the user's local filesystem. No unauthorized network transmission of user data or secrets was found.
[PROMPT_INJECTION]: The skill processes untrusted external data (article HTML). It mitigates potential indirect prompt injection by converting HTML to Markdown using a structured parser and wrapping the output in defined Markdown containers.
Ingestion points: WeChat API responses and public web pages fetched in fetch_api.py, fetch_cookie.py, and fetch_mp.py.
Boundary markers: Articles are saved with YAML frontmatter and Markdown headers as defined in save_articles.py.
Capability inventory: Local file system writes and network GET/POST requests using the Python standard library.
Sanitization: Uses html.parser to convert HTML to Markdown, effectively stripping or neutralizing executable script tags.

soia-pkm-clip-wechat-account