pptx
Fail
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The script scripts/office/soffice.py dynamically writes C source code to the file system, compiles it using gcc at runtime, and uses the LD_PRELOAD environment variable to inject the resulting shared library into subprocesses. This technique, while intended to bypass socket restrictions, represents a high-risk security pattern.
- [COMMAND_EXECUTION]: Multiple Python scripts (scripts/thumbnail.py, scripts/office/soffice.py, scripts/office/validators/redlining.py) use the subprocess module to execute system commands, including soffice for PDF conversion, pdftoppm for image generation, git for diffing, and gcc for compilation.
- [EXTERNAL_DOWNLOADS]: The skill requires several external dependencies to be installed from public registries, including markitdown and Pillow via pip, and pptxgenjs via npm.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8) because it extracts text from untrusted user-provided PowerPoint files and incorporates it into the agent's context. The skill lacks explicit boundary markers or sanitization for this extracted text while maintaining access to high-privilege capabilities like subprocess execution and file system writes.
Recommendations
- AI detected serious security threats
Audit Metadata