Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates PDF manipulation by instructing the agent to execute various bundled Python scripts (e.g.,
scripts/extract_form_field_info.py,scripts/fill_fillable_fields.py) from the command line. This is the intended design for providing programmatic PDF processing capabilities. - [EXTERNAL_DOWNLOADS]: Documentation within the skill identifies several third-party libraries (such as
pypdf,pdfplumber, andreportlab) to be installed from official registries to support advanced PDF features. - [DYNAMIC_EXECUTION]: The
scripts/fill_fillable_fields.pyfile implements a runtime monkeypatch for thepypdflibrary to fix an upstream bug in selection list handling. This is a specific workaround required for the skill's form-filling functionality. - [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted PDF data, which is a surface for indirect prompt injection. Ingestion points: PDF files are read using
pypdf.PdfReaderin multiple scripts. Boundary markers: None are defined in the scripts or instructions. Capability inventory: The skill has the ability to read and write files and execute bundled Python scripts. Sanitization: No content sanitization is applied to the data extracted from PDF files.
Audit Metadata