The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits search queries and a user-provided API key to redfox.hk. This domain is an official resource belonging to the skill's author, used to provide the intended data service.
[CREDENTIALS_UNSAFE]: The script scripts/search_work.py includes logic to search for the REDFOX_API_KEY within sensitive local shell configuration files, including ~/.zshrc, ~/.bashrc, ~/.bash_profile, and ~/.profile, if it is not found in the environment variables.
[INDIRECT_PROMPT_INJECTION]: The skill processes content fetched from an external API, creating a potential surface for indirect injection.
Ingestion points: The search results (titles, authors, etc.) are fetched via the post_redfox function in scripts/search_work.py.
Boundary markers: External data is strictly confined to a Markdown table structure during rendering.
Capability inventory: The skill is limited to performing network requests and reading local configuration files for keys; it does not execute code or write files based on the ingested API data.
Sanitization: The escape_md function in the search script sanitizes incoming data by removing newlines and escaping pipe characters to prevent Markdown table manipulation.
[EXTERNAL_DOWNLOADS]: The skill establishes network connections to redfox.hk for data retrieval. Notably, the script explicitly disables SSL hostname verification (context.check_hostname = False) for this connection to ensure compatibility with the host, which reduces protection against man-in-the-middle attacks.

dongchedi-work-search