skills/isomoes/skills/pdf/Gen Agent Trust Hub

pdf

Warn

Audited by Gen Agent Trust Hub on Apr 14, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The script scripts/fill_fillable_fields.py implements a monkeypatching technique to modify the pypdf library at runtime. Specifically, it redefines the DictionaryObject.get_inherited method to work around a bug in the library's handling of selection list fields. Runtime code alteration is a form of dynamic execution.
  • [COMMAND_EXECUTION]: The skill provides instructions and examples for executing various system-level command-line utilities, including pdftotext, qpdf, pdftk, and pdftoppm. These tools interact with the host system to perform PDF conversions and manipulations.
  • [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because its primary function involves processing data from untrusted PDF files.
  • Ingestion points: PDF content is read and parsed using pypdf and pdfplumber across multiple scripts and documentation files.
  • Boundary markers: The instructions do not define security delimiters or include warnings for the agent to ignore potentially malicious commands embedded within extracted PDF text.
  • Capability inventory: The skill has significant capabilities, including the ability to write files to the local filesystem and execute shell commands.
  • Sanitization: There is no evidence of sanitization or validation of content extracted from PDFs before it enters the agent's context.
  • [PROMPT_INJECTION]: There is a potential metadata inconsistency regarding the skill's origin. The skill is attributed to the author isomoes, but the LICENSE.txt file contains a copyright notice for Anthropic, PBC with highly restrictive proprietary terms. This discrepancy in provenance and licensing information can be misleading regarding the ownership and safety of the code.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 14, 2026, 01:25 PM
Security Audit — agent-trust-hub — pdf