The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess calls to execute several well-known command-line utilities, including qpdf for document manipulation, pdftotext and pdfimages from poppler-utils, and magick for image processing.
[EXTERNAL_DOWNLOADS]: The skill's documentation recommends the use of widely-used third-party libraries from official package registries, such as pypdf, pdfplumber, reportlab, pdf2image, pytesseract, pypdfium2, and pdf-lib.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PDF content (via text extraction and OCR). Because this external data is incorporated into the agent's context without sanitization or boundary markers, it presents a surface for potential indirect prompt injection attacks.
Ingestion points: PDF files processed by pypdf, pdfplumber, pdftotext, and pytesseract as described in SKILL.md, forms.md, and reference.md.
Boundary markers: No explicit instructions are provided to the agent to treat the extracted content as untrusted data or to use delimiters.
Capability inventory: The skill has the ability to execute shell commands and perform file system writes.
Sanitization: The skill does not include automated sanitization of the data extracted from PDF documents before processing.
[DYNAMIC_EXECUTION]: The file scripts/fill_fillable_fields.py performs a runtime modification (monkeypatching) of the pypdf library's DictionaryObject.get_inherited method to adjust its behavior for specific PDF field structures.

pdf