The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's instructions in forms.md and SKILL.md guide the agent to use local command-line utilities, including qpdf, pdftk, and ImageMagick (magick), to perform PDF and image manipulation tasks.
[INDIRECT_PROMPT_INJECTION]: The skill processes external PDF files, creating an attack surface where malicious content within a PDF could potentially influence agent behavior. If a PDF contains text designed to look like instructions, an agent extracting that text might be susceptible to indirect injection.
Ingestion points: scripts/extract_form_structure.py, scripts/extract_form_field_info.py, and Python code snippets in SKILL.md that extract text and metadata from PDF files.
Boundary markers: The output of the extraction scripts is provided as plain text or JSON without specific delimiters or warnings to the agent to disregard instructions within the extracted content.
Capability inventory: The skill includes the ability to write files (PDFs, images, JSON) and execute local system commands for processing.
Sanitization: The skill does not perform sanitization or filtering on the text content extracted from user-supplied PDFs.
[DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py implements a runtime monkeypatch to the pypdf.generic.DictionaryObject.get_inherited method to adjust the library's behavior when handling specific PDF field dictionary attributes.
[EXTERNAL_DOWNLOADS]: The documentation lists several external dependencies, including Python libraries (pypdf, pdfplumber, reportlab, pytesseract, pdf2image, pypdfium2) and Node.js libraries (pdf-lib, pdfjs-dist), as well as system tools (poppler-utils, qpdf).

pdf