book-to-skill

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill operates locally and focuses on transforming user-provided documents into structured markdown files.
  • [SAFE]: Extraction logic (in book_to_skill/sanitize.py) proactively identifies and strips zero-width characters, bidirectional overrides, and Unicode tags to prevent "document-borne" prompt injection from untrusted source files.
  • [SAFE]: The instructions in SKILL.md (Step 9.5) and the tool tools/scan_generated_skill.py implement an automated security audit that flags instruction-override phrases, model control tags, and exfiltration-shaped language in the generated skill files.
  • [SAFE]: Subprocess calls for document extractors (e.g., pdftotext, ebook-convert) use absolute paths and validated commands to prevent argument injection.
  • [SAFE]: The project demonstrates high security maturity by maintaining a SECURITY-NOTICE.md to warn users about malicious third-party re-uploads and providing a clear vulnerability reporting policy.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 12:25 AM
Security Audit — agent-trust-hub — book-to-skill