Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines workflows to ingest and process external PDF documents, which could contain instructions intended to influence the agent's behavior during extraction or form filling.
- Ingestion points:
scripts/extract_form_field_info.pyandscripts/convert_pdf_to_images.pyparse external PDF structures and content. - Boundary markers: The
forms.mdfile contains explicit instructions and "CRITICAL" warnings for the agent to follow a strict order of operations and perform visual validation. - Capability inventory: The skill has the capability to extract text/tables, convert pages to images, and write new or modified PDF files to the local file system.
- Sanitization:
scripts/fill_fillable_fields.pyvalidates requested field IDs and values against a JSON schema extracted from the PDF before applying changes. - [DYNAMIC_EXECUTION]: One of the included scripts modifies the behavior of the
pypdflibrary at runtime. - Evidence: In
scripts/fill_fillable_fields.py, themonkeypatch_pydpf_methodfunction redefinesDictionaryObject.get_inheritedfrom thepypdfpackage during execution. - Context: This is a localized runtime modification used to fix a specific bug in
pypdfrelated to selection list field values and does not involve executing code derived from untrusted sources.
Audit Metadata