Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to process external PDF documents using libraries like
pypdf,pdfplumber, andpytesseract. While essential for the skill's functionality, ingesting untrusted content creates a potential surface for indirect prompt injection, where hidden text or instructions within a document could attempt to influence the agent's behavior. - Ingestion points:
PdfReader.pages.extract_text()inSKILL.md,pdfplumber.open()inSKILL.md, andpytesseract.image_to_string()inSKILL.md. - Boundary markers: None identified in the provided instructions.
- Capability inventory: File system write operations (
merged.pdf,report.pdf,extracted_tables.xlsx) and shell command execution for PDF utilities. - Sanitization: No explicit sanitization of extracted text is described before it is processed by the agent.
- Dynamic Code Modification (Monkeypatching): The script
scripts/fill_fillable_fields.pycontains a functionmonkeypatch_pydpf_methodthat overrides a method in thepypdflibrary at runtime. This modification is used to address a specific bug in the library's handling of selection lists. While this involves dynamic execution, the logic is constrained, transparent, and used solely to ensure correct functionality of the form-filling process. - Command-Line Tool Execution: The skill provides instructions for using external utilities such as
pdftotext,qpdf, andpdftk. These tools are standard for PDF processing in development environments and are used here within their intended scope for document manipulation.
Audit Metadata