The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external PDF files, which are untrusted sources.
Ingestion points: Multiple methods in SKILL.md (extracting text/tables via pypdf and pdfplumber) and scripts/extract_form_structure.py (extracting text labels for field identification) read content directly from user-provided PDF files.
Boundary markers: The instructions do not include specific boundary markers or instructions for the agent to ignore potentially malicious commands embedded within the extracted PDF text.
Capability inventory: The skill includes scripts that can write to the local filesystem (scripts/fill_fillable_fields.py, scripts/fill_pdf_form_with_annotations.py) and documentation suggesting the execution of CLI tools like qpdf and pdftk.
Sanitization: There is no evidence of sanitization, filtering, or validation performed on the text extracted from the PDFs before it is presented to the agent context.
[DYNAMIC_EXECUTION]: One of the provided scripts modifies library behavior at runtime.
Evidence: scripts/fill_fillable_fields.py contains a function monkeypatch_pydpf_method that replaces the get_inherited method of pypdf.generic.DictionaryObject with a custom implementation to handle specific PDF field attributes. While used for functional compatibility, this is a form of runtime code modification.

pdf