The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided PDF files through various extraction and manipulation scripts. This creates a surface for indirect prompt injection where malicious instructions could be embedded in PDF text, metadata, or form fields to influence the agent's behavior.
Ingestion points: The skill reads external PDF files in forms.md (via check_fillable_fields, extract_form_field_info.py, convert_pdf_to_images.py) and SKILL.md.
Boundary markers: There are no explicit instructions to the agent to ignore or delimit potentially untrusted content within the processed PDFs.
Capability inventory: The skill possesses the capability to write files (pypdf's PdfWriter.write) and execute shell commands via command-line tools like pdftotext and qpdf.
Sanitization: No sanitization or filtering of the content extracted from the PDFs is performed before it is processed by the agent.
[DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py utilizes runtime monkeypatching to address a known bug in the pypdf library regarding the retrieval of inherited field attributes. While this is a targeted fix for a specific library issue, it represents a form of self-modifying code at runtime.
Evidence: The monkeypatch_pydpf_method function in scripts/fill_fillable_fields.py (lines 78-95) dynamically replaces DictionaryObject.get_inherited with a custom implementation.

pdf