The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py performs monkeypatching on the pypdf library at runtime. Specifically, it overrides the get_inherited method of pypdf.generic.DictionaryObject to address a bug in the library's handling of selection list fields. This dynamic modification of an imported library is a notable technique, though here it is used for a documented bug fix.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external PDF documents, creating a potential surface for indirect prompt injection attacks.
Ingestion points: Data is read from PDFs using libraries such as pypdf and pdfplumber, and command-line tools like pdftotext, as seen in SKILL.md and scripts/extract_form_field_info.py.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the extracted PDF text to protect the agent's context.
Capability inventory: The skill possesses file system write access (PdfWriter.write()) and the ability to execute command-line utilities, which could be leveraged if an agent obeys instructions hidden in a document.
Sanitization: Extracted content is provided to the agent without evidence of explicit sanitization or filtering.
[COMMAND_EXECUTION]: The skill documentation and scripts facilitate the execution of several command-line utilities for PDF manipulation, including qpdf, pdftotext, pdftk, and pdftoppm. These are standard tools used for merges, splits, text extraction, and document rendering.
[EXTERNAL_DOWNLOADS]: The documentation references and recommends the installation of several well-known third-party libraries to provide its functionality, including pytesseract, pdf2image, pypdfium2, pdf-lib, and pdfjs-dist.

pdf