The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python scripts (fetch_api.py, fetch_cookie.py, fetch_mp.py) to perform network requests and write processed articles to the local file system (Obsidian vault). These actions are consistent with the skill's stated purpose of archiving content.
[EXTERNAL_DOWNLOADS]: The scripts fetch article metadata and HTML content from WeChat servers (api.weixin.qq.com, mp.weixin.qq.com). All network operations target legitimate service endpoints belonging to the platform being archived.
[CREDENTIALS_UNSAFE]: The skill requires sensitive information such as WECHAT_APP_SECRET and WECHAT_COOKIE. The implementation follows security best practices by providing a config.example.yml with placeholders and instructing users to store actual secrets in a private configuration file (~/.config/soia-skills/.../config.yml) that is not part of the vault or code repository.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (WeChat article content) and land it in the agent's context (the Obsidian vault). This creates an ingestion surface for potential indirect prompt injection attacks where malicious instructions hidden in a WeChat article could influence the agent's behavior during future interactions.
Ingestion points: Content is ingested through the fetch_article_body function in scripts/fetch_cookie.py and scripts/fetch_mp.py, and _news_item_to_article in scripts/fetch_api.py.
Boundary markers: The resulting Markdown files do not appear to use explicit boundary markers or warnings to the agent regarding the untrusted nature of the archived text.
Capability inventory: The skill has the capability to write files to the vault (scripts/save_articles.py) and make network requests via urllib.request.
Sanitization: The html_to_markdown function in scripts/save_articles.py provides basic sanitization by explicitly skipping <script> and <style> tags during HTML-to-Markdown conversion.

soia-pkm-clip-gzh