The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external PDF documents, creating an attack surface for indirect prompt injection.
Ingestion points: The scripts scripts/extract_form_structure.py, scripts/extract_form_field_info.py, and scripts/convert_pdf_to_images.py read user-provided PDF files for analysis.
Boundary markers: The skill does not implement delimiters or explicit instructions for the agent to ignore embedded commands within the PDF content during extraction.
Capability inventory: The skill can execute shell commands (qpdf, pdftotext, magick) and write files, which could be abused if the agent follows instructions found inside a processed PDF.
Sanitization: No sanitization or validation of the extracted PDF text or structure is performed before presenting it to the agent for decision-making.
[COMMAND_EXECUTION]: The skill instructions and scripts rely on several external command-line utilities for core functionality.
Evidence: Documentation in SKILL.md and FORMS.md guides the agent to use pdftotext, qpdf, pdftk, and magick (ImageMagick) for text extraction, merging, and image manipulation.
[REMOTE_CODE_EXECUTION]: The skill uses dynamic execution techniques to modify library behavior at runtime.
Evidence: The script scripts/fill_fillable_fields.py performs monkeypatching of the pypdf library's DictionaryObject.get_inherited method to handle specific PDF choice field attributes.
[EXTERNAL_DOWNLOADS]: The skill utilizes several standard Python and Node.js packages for PDF processing.
Evidence: Packages such as pypdf, pdfplumber, reportlab, pdf-lib, and pdfjs-dist are required by the provided scripts or suggested in the documentation.

pdf