The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external, untrusted PDF files to extract text and structure, creating a surface for indirect prompt injection. * Ingestion points: Text and structure extraction logic in SKILL.md and scripts/extract_form_structure.py. * Boundary markers: The skill does not implement delimiters or warnings to ignore instructions embedded within the processed PDF data. * Capability inventory: The skill has capabilities for local command execution (magick, qpdf, etc.) and file system writes. * Sanitization: There is no evidence of sanitization or validation of the extracted data before it is utilized.
[COMMAND_EXECUTION]: The instructions direct the agent to run various local command-line tools and Python scripts to perform PDF operations. * Evidence: SKILL.md and forms.md provide commands for tools such as pdftotext, qpdf, and magick. * Evidence: Local Python scripts in the scripts/ directory are executed via the command line to process files.
[COMMAND_EXECUTION]: A provided script performs dynamic modification of an external library at runtime. * Evidence: scripts/fill_fillable_fields.py contains a monkeypatch that replaces the DictionaryObject.get_inherited method in the pypdf library to handle specific attribute types.

pdf