doc-to-vector-dataset-generator

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external document content which represents a surface for indirect prompt injection attacks against downstream AI systems consuming the generated datasets.
  • Ingestion points: The extract_markdown and extract_pdf functions ingest raw text content from files located in the input_dir path.
  • Boundary markers: The current implementation does not define specific delimiters or "ignore" instructions within the exported JSONL chunks to help downstream agents distinguish data from instructions.
  • Capability inventory: The skill utilizes filesystem read/write capabilities (open, glob) to facilitate document processing and dataset export.
  • Sanitization: Content cleaning is restricted to whitespace normalization and URL removal, but does not include sanitization or filtering for embedded prompt injection payloads.
  • [EXTERNAL_DOWNLOADS]: The skill relies on well-known external libraries to perform its core functions.
  • Dependencies: Requires the installation of pymupdf (for PDF text extraction) and scikit-learn (for TF-IDF vectorization and cosine similarity deduplication).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 09:30 AM
Security Audit — agent-trust-hub — doc-to-vector-dataset-generator