Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The script
scripts/fill_fillable_fields.pyperforms a monkeypatch at runtime, replacing theDictionaryObject.get_inheritedmethod in thepypdflibrary with a custom implementation (monkeypatch_pydpf_method). This is documented as a workaround for a specific bug inpypdfversion 5.7.0 that causes crashes when processing choice list fields. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process untrusted PDF documents, which presents a surface for indirect prompt injection if malicious content within a PDF influences the agent's logic.
- Ingestion points: Untrusted data enters via PDF files read by
pypdfandpdfplumber(as seen inscripts/extract_form_field_info.pyandSKILL.md) and configuration files loaded viajson.load(). - Boundary markers: The skill documentation in
forms.mdexplicitly mandates a multi-step verification process, requiring the agent to perform visual analysis of images and run automated checks (scripts/check_bounding_boxes.py) before completing tasks. - Capability inventory: The skill possesses the capability to write PDF files (
writer.write), generate images (image.save), and execute command-line utilities such asqpdfandpdftotext. - Sanitization: The
scripts/fill_fillable_fields.pyutility includes avalidation_error_for_field_valuefunction that strictly validates input values against the PDF's schema (e.g., checking valid options for checkboxes and radio groups) before writing data.
Audit Metadata