The Agent Skills Directory

[SAFE]: No malicious activities such as data exfiltration, credential harvesting, or prompt injection were detected. The skill's functionality is consistent with its stated purpose of PDF document manipulation.
[DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py performs a localized monkeypatch of the pypdf library to resolve a known formatting issue with selection list fields. This modification is hardcoded and used solely for functional correctness during the form-filling process.
[INDIRECT_PROMPT_INJECTION]: The skill processes external PDF files, creating a standard attack surface for indirect prompt injection. The skill mitigates this through structured instructions that require the agent to perform validation and human review before completing document transformations.
Ingestion points: PDF documents are analyzed by scripts/extract_form_field_info.py and scripts/check_fillable_fields.py to extract structure and metadata.
Boundary markers: The forms.md file provides a strict, sequential workflow (e.g., 'CRITICAL: You MUST complete these steps in order') to ensure the agent processes data predictably.
Capability inventory: The skill includes several Python scripts that read and write PDF, JSON, and PNG files to the local filesystem.
Sanitization: scripts/fill_fillable_fields.py performs validation of field IDs and data types against the source PDF's extracted field information before writing data.

pdf