The Agent Skills Directory

Command Execution: The skill provides detailed instructions and examples for using various external command-line utilities such as qpdf, pdftotext, and ImageMagick (magick). While these are standard tools for PDF and image processing workflows, their use involves executing commands on the underlying system to manipulate files.
External Dependencies: The skill relies on a variety of third-party Python and JavaScript libraries (e.g., pypdf, pdfplumber, reportlab, pdf-lib). These are established packages in the PDF processing ecosystem, but users should ensure they are installed from trusted official registries.
Dynamic Execution: The script scripts/fill_fillable_fields.py utilizes monkeypatching to modify the internal behavior of the pypdf library at runtime. This technique is specifically implemented to improve the handling of inherited field options in PDF forms but represents a dynamic alteration of third-party code during execution.
Indirect Prompt Injection Surface: The skill is designed to ingest and process data from external PDF files, creating a potential surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context through PDF files processed by scripts in the scripts/ directory, which extract text, metadata, and form structures.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the content extracted from the PDFs.
Capability inventory: The skill possesses the capability to execute shell commands (qpdf, pdftotext, magick), write files to the local system (PDF, JSON, PNG, TXT, XLSX), and perform complex PDF manipulations.
Sanitization: The provided scripts do not appear to implement explicit sanitization or filtering of the text or metadata extracted from source PDF files before it is processed by the agent.

pdf