The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands including curl, cat, and python3 (for JSON parsing) to perform its core functions such as searching, browsing, and downloading books.
[DATA_EXPOSURE]: The skill is designed to read a local configuration file at ~/.config/calibre-library/config.json to obtain the user's base_url, username, and password. While storing secrets in a local configuration file is a standard management practice, it involves the agent handling sensitive credentials which are subsequently used in network requests via Basic Auth.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It retrieves data from an external Calibre API, including book titles, author names, and HTML formatted comments, which are then integrated into the agent's context and displayed to the user.
Ingestion points: Data enters the agent context through the Calibre AJAX API endpoints, specifically ajax/books and ajax/search.
Boundary markers: The instructions do not specify the use of delimiters or protective wrappers (e.g., "ignore embedded instructions") when presenting book metadata or descriptions to the user.
Capability inventory: The skill possesses the ability to execute shell commands, read local files, and download files to the user's system.
Sanitization: There is no evidence of sanitization or filtering of the HTML content received from the comments field before it is processed or displayed.

calibre-library