The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PDF documents through various scripts such as extract_form_field_info.py and fill_fillable_fields.py. This creates an attack surface where instructions embedded within a processed PDF could attempt to influence the agent's logic.
Ingestion points: Untrusted data enters the system via scripts/extract_form_field_info.py, scripts/fill_fillable_fields.py, and scripts/check_fillable_fields.py which parse PDF content.
Boundary markers: The skill implements a structured, multi-step workflow in forms.md that requires specific JSON formatting and manual validation steps, providing a functional boundary between data extraction and action.
Capability inventory: The skill possesses capabilities for file system operations (reading/writing PDFs, JSON, and PNGs) and executing external CLI tools like qpdf and pdftotext.
Sanitization: The scripts rely on standard parsing behavior from the pypdf and json libraries without additional content-level sanitization.
[DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py contains a monkeypatch_pydpf_method function that dynamically overrides pypdf.generic.DictionaryObject.get_inherited at runtime. This is a localized modification designed to fix a specific known bug in the pypdf library (version 5.7.0) regarding selection list fields.

pdf