The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PDF files that could contain malicious instructions designed to influence the agent.\n
Ingestion points: PDFs are processed by scripts/extract_form_field_info.py, scripts/check_fillable_fields.py, and referenced in SKILL.md via pypdf and pdfplumber.\n
Boundary markers: No explicit delimiters or instructions to ignore embedded content are used when extracting text for the agent.\n
Capability inventory: The skill can execute various subprocesses (e.g., qpdf, pdftotext, pdftk) and has extensive read/write access to the local file system.\n
Sanitization: Extracted content is passed to the agent without sanitization or filtering.\n- [DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py modifies the behavior of the pypdf library at runtime.\n
Evidence: The monkeypatch_pydpf_method function replaces pypdf.generic.DictionaryObject.get_inherited with a custom implementation to address a bug in the library's handling of selection lists.\n- [COMMAND_EXECUTION]: The skill relies on several external command-line utilities for core functionality.\n
Evidence: SKILL.md and reference.md instruct the agent to use qpdf, pdftotext, pdfimages, pdftoppm, and pdftk for tasks like merging, splitting, and text extraction.

pdf