The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes PDF documents from external and untrusted sources, which serves as a surface for indirect prompt injection attacks.
Ingestion points: Data is extracted from PDF files using libraries like pypdf, pdfplumber, and pytesseract (OCR) as described in SKILL.md and helper scripts.
Capability inventory: The skill has the ability to read and write files to the filesystem and execute shell commands via provided instructions.
Boundary markers: The instructions lack explicit directives for the agent to treat extracted PDF text as untrusted or to ignore instructions embedded within document content.
Sanitization: No sanitization or validation of extracted text is implemented before processing by the agent.
[DYNAMIC_EXECUTION]: The skill performs dynamic code modification of a dependency at runtime.
Evidence: scripts/fill_fillable_fields.py implements a monkeypatch on the pypdf.generic.DictionaryObject.get_inherited method to correct specific behavior related to selection list fields in PDF forms.
[COMMAND_EXECUTION]: The skill documentation provides instructions for the agent to execute various command-line operations.
Evidence: SKILL.md and forms.md guide the agent to use CLI tools such as pdftotext, qpdf, pdfimages, and the included custom Python scripts for document manipulation.

pdf