office-academic-skill
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill extracts and processes text from untrusted academic documents (PDFs and LaTeX files) to generate presentations and reports.
- Ingestion points: The
references/thesis-defense-pptx/scripts/extract_thesis_context.pyscript uses PyMuPDF and pypdf to read external document content. - Boundary markers: Extraction scripts do not consistently apply delimiters or instructions to ignore embedded commands when inserting untrusted text into the agent's context.
- Capability inventory: The skill has extensive file system access, document manipulation capabilities, and the ability to execute local commands.
- Sanitization: No specific filtering or sanitization of the extracted text is performed before it is interpolated into prompts.
- [COMMAND_EXECUTION]: Several Python and PowerShell scripts execute external binaries and system tools to perform document processing tasks.
- The
pack.pyscript in both Word and PowerPoint modules executessoffice(LibreOffice) to validate document integrity. - The
thumbnail.pyscript executessofficefor PDF conversion andpdftoppmfor slide image generation. - The
redlining.pyscript executesgit diffto provide character-level comparison between document versions. - [DYNAMIC_EXECUTION]: The skill uses automation technologies that execute code or content at runtime to generate documents.
- The
references/office-pptx/scripts/html2pptx.jsscript uses the Playwright library to launch a Chromium browser and render HTML templates. - PowerShell scripts located in
references/thesis-defense-pptx/scripts/utilize COM objects to automate Microsoft PowerPoint for slide cloning and export.
Audit Metadata