The Agent Skills Directory

PROMPT_INJECTION (HIGH): The skill is vulnerable to Indirect Prompt Injection (IPI). It ingests untrusted PDF data which is subsequently used to influence agent decisions during form-filling and extraction tasks. A malicious PDF could embed instructions to manipulate the agent's logic or output.
Ingestion points: scripts/extract_form_field_info.py and scripts/convert_pdf_to_images.py.
Boundary markers: Lacks semantic delimiters for extracted untrusted content.
Capability inventory: File system modification via pypdf, pdfplumber, and reportlab.
Sanitization: Only spatial bounding box validation is present (scripts/check_bounding_boxes.py); no content-based sanitization or instruction-ignoring markers are applied.
COMMAND_EXECUTION (MEDIUM): The script scripts/fill_fillable_fields.py utilizes runtime monkeypatching on the pypdf.generic.DictionaryObject class to work around a selection list bug. This dynamic modification of library behavior at runtime is a risk-sensitive execution pattern.

pdf