convert-pdf-to-epub
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: CRITICALCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- COMMAND_EXECUTION (SAFE): The skill provides several CLI tools (convert.py, analyze.py, validate.py) for the conversion workflow. These use the standard argparse library and implement safe local file path handling without invoking shell subprocesses or arbitrary commands.\n- EXTERNAL_DOWNLOADS (SAFE): Python dependencies in requirements.txt (such as pymupdf, beautifulsoup4, and Pillow) are reputable and appropriate for the skill's stated purpose. Documentation links point exclusively to trusted Anthropic and GitHub domains.\n- DATA_EXFILTRATION (SAFE): There are no network requests, API calls, or attempts to access sensitive system files (e.g., ~/.ssh or .env). The processing is entirely local.\n- Indirect Prompt Injection (LOW): As a document conversion tool, the skill has an inherent surface for indirect prompt injection by processing untrusted PDF data. If a PDF contains malicious instructions, they could be carried over to the EPUB and influence a downstream AI agent.\n
- Ingestion points: conversion/pdf_analyzer.py and conversion/strategies/base_strategy.py read content from external PDF files.\n
- Boundary markers: Absent. Extracted text is not wrapped in protective delimiters or warnings.\n
- Capability inventory: File write access via core/epub_builder.py and image optimization via core/image_optimizer.py.\n
- Sanitization: Limited to HTML escaping in conversion/detectors/endnote_formatter.py; no general instruction sanitization is performed on extracted text.
Recommendations
- Contains 1 malicious URL(s) - DO NOT USE
Audit Metadata