The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PDF documents and derived images to extract text, tables, and form field information. This creates a surface where malicious content within a document could attempt to influence the agent's interpretation of data or its subsequent actions.
Ingestion points: Text and table extraction methods described in SKILL.md and visual analysis of PNG images described in forms.md.
Boundary markers: No explicit delimiters or instructions are used to separate user-provided PDF content from the agent's instruction context.
Capability inventory: The skill enables writing to the local file system (PDFs, Excel, images) and executing PDF utility commands (e.g., qpdf, pdftoppm).
Sanitization: Extracted content is processed as raw data without specific filtering for malicious instructions.
[DYNAMIC_EXECUTION]: The script scripts/fill_fillable_fields.py implements a runtime monkeypatch for the pypdf library's DictionaryObject.get_inherited method. This is used to resolve a known incompatibility in the library when processing selection lists. The implementation is static and scoped to improving the reliability of form-filling operations.

pdf