gutenberg

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads book metadata and ebook files from gutendex.com and gutenberg.org. These are established, well-known services for accessing Project Gutenberg's catalog.
  • [COMMAND_EXECUTION]: The skill includes a portable Python CLI (scripts/gutenberg) that performs network requests via urllib and writes downloaded book content to the local filesystem.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes text from external books. While these are public-domain literary works, any skill that feeds untrusted external data into an agent's context presents a theoretical surface for indirect prompt injection if the text contains instructions the agent might follow.
  • Ingestion points: scripts/gutenberg fetches data from gutendex.com (API) and gutenberg.org (Book text/EPUB).
  • Boundary markers: The script performs boilerplate stripping but does not include explicit delimiters or 'ignore instructions' wrappers for the extracted text.
  • Capability inventory: The script performs file writes (open(...).write()) and network GET requests (urllib.request.urlopen).
  • Sanitization: The script performs basic HTML stripping for EPUB files and regex-based boilerplate removal, but no semantic sanitization of the book content itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 05:57 PM
Security Audit — agent-trust-hub — gutenberg