The Agent Skills Directory

[PROMPT_INJECTION]: The skill downloads and processes text content from Project Gutenberg, an external public source. If a downloaded book contains malicious instructions (e.g., "Ignore all previous instructions"), it could lead to indirect prompt injection.
Ingestion points: Book content and metadata are retrieved from gutenberg.org and gutendex.com via the scripts/gutenberg tool.
Boundary markers: The skill does not explicitly define delimiters to separate untrusted book content from agent instructions.
Capability inventory: The tool performs network read operations and local file write operations across its functions in scripts/gutenberg.
Sanitization: The script strips licensing boilerplate but does not filter the literary content for potential prompt injection patterns.
[EXTERNAL_DOWNLOADS]: The CLI fetches book metadata and content from the official Project Gutenberg website and the Gutendex API.
The script uses urllib.request to communicate with https://gutendex.com and https://www.gutenberg.org using a custom User-Agent.
[COMMAND_EXECUTION]: The skill relies on executing the provided Python script scripts/gutenberg to perform search, download, and extraction tasks.
The script uses Python standard library modules (such as urllib, zipfile, and html.parser) to manage its operations safely without external dependencies.

gutenberg