The Agent Skills Directory

[SAFE]: No malicious behavior, obfuscation, or unauthorized data access was detected. The skill is authored by a trusted entity and uses legitimate dependencies.
[PROMPT_INJECTION]: The skill processes untrusted PDF documents, presenting an inherent surface for indirect prompt injection. 1. Ingestion points: pypdf is used for form extraction and pdfplumber is used for text/table parsing. 2. Boundary markers: forms.md contains strict workflow instructions ('CRITICAL: You MUST complete these steps in order') to constrain agent behavior. 3. Capability inventory: The skill can write files to the local disk and execute subprocesses for PDF tools. 4. Sanitization: scripts/check_bounding_boxes.py and scripts/fill_fillable_fields.py provide structural validation and verification of field IDs/values, although no text content filtering is implemented. This surface is considered a standard operational risk for document processing tools.
[COMMAND_EXECUTION]: The skill provides instructions for using standard PDF utilities (qpdf, pdftotext, pdftk) on local files, which is appropriate for its primary function.
[REMOTE_CODE_EXECUTION]: scripts/fill_fillable_fields.py applies a runtime monkeypatch to the pypdf library to resolve a specific selection list bug. This is a static code modification for compatibility and does not facilitate the execution of untrusted external code.

pdf