The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by processing untrusted data from PDF files.\n
Ingestion points: PDF content and form field metadata are read in scripts/extract_form_field_info.py, scripts/convert_pdf_to_images.py, and various code snippets in SKILL.md and reference.md.\n
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are implemented when extracting or processing PDF text.\n
Capability inventory: The skill includes scripts for file-writing (fill_fillable_fields.py, fill_pdf_form_with_annotations.py) and suggests executing multiple command-line utilities.\n
Sanitization: No sanitization or validation of content extracted from PDFs is performed before it is used by the agent.\n- [COMMAND_EXECUTION]: The skill's operation relies on the execution of local scripts and system-level binaries.\n
Script execution: The instructions in forms.md direct the agent to run several local Python scripts to process PDFs.\n
External tools: Recommends using command-line tools such as qpdf, pdftotext, pdftk, and pdftoppm.\n
Dynamic behavior: scripts/fill_fillable_fields.py uses a runtime monkeypatch to modify the pypdf library's internal method get_inherited to resolve a bug.\n- [EXTERNAL_DOWNLOADS]: The documentation suggests installing multiple third-party libraries from well-known sources.\n
Python packages: Recommends pypdf, pdfplumber, reportlab, pytesseract, pdf2image, pandas, and pypdfium2.\n
Node.js packages: References pdf-lib and pdfjs-dist in the advanced guide.

pdf