skills/anthropics/skills/pdf/Gen Agent Trust Hub

pdf

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill is designed to ingest and process data from untrusted PDF files.
  • Ingestion points: PDF extraction in scripts/extract_form_structure.py and scripts/extract_form_field_info.py.
  • Boundary markers: Delimiters and instructions to ignore embedded commands are absent.
  • Capability inventory: File system write access via pypdf, pdfplumber, and Pillow; execution of CLI tools (qpdf, pdftotext) is recommended in documentation.
  • Sanitization: No sanitization of extracted text is performed before it is passed to the agent context.
  • [Dynamic Execution] (LOW): The script scripts/fill_fillable_fields.py uses monkeypatching to modify the pypdf library's DictionaryObject.get_inherited method at runtime to correctly handle specific form attributes.
  • [External Downloads] (LOW): The skill documentation recommends unversioned installation of several Python packages (pytesseract, pdf2image) and system-level utilities (poppler-utils, qpdf, pdftk). This finding is downgraded from MEDIUM because the skill is attributed to a trusted organization (Anthropic).
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:51 PM