The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to ingest and process untrusted PDF documents, which serves as a potential surface for indirect prompt injection. \n
Ingestion points: scripts/extract_form_field_info.py, scripts/check_fillable_fields.py, and SKILL.md (via extraction examples). \n
Boundary markers: The skill relies on procedural boundaries described in forms.md (manual 'Visual Analysis' steps) rather than automated sanitization. \n
Capability inventory: Local file creation and PDF annotation via pypdf and PIL; documentation references CLI utilities (qpdf, pdftk). No network capabilities were found. \n
Sanitization: Relies on standard parsing via pypdf and pdfplumber without additional sanitization of extracted text content. \n- [Dynamic Execution] (LOW): The script scripts/fill_fillable_fields.py performs a runtime monkeypatch on the pypdf library to address a specific bug in selection list handling. This is a localized and documented modification of library behavior. \n- [Command Execution] (LOW): Documentation and code within SKILL.md and the toolkit reference the use of standard external command-line utilities such as qpdf, pdftk, and poppler-utils. These are standard tools for this domain but involve subprocess spawning with local file paths.

pdf