academic-presentations
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the
nanobananaextension from an external GitHub repository (https://github.com/gemini-cli-extensions/nanobanana). This introduces third-party code from a source outside the verified environment. - [COMMAND_EXECUTION]: The workflow relies on executing shell commands for document conversion using
LibreOffice(soffice) andpoppler(pdftoppm), as well as media processing withffmpegand image generation via thegeminiCLI. - [INDIRECT_PROMPT_INJECTION]: The skill processes research papers and scripts to generate slide content and narration, which could allow malicious instructions embedded in the input documents to influence the agent's output.
- Ingestion points: External research papers (PDF/text) and talking points provided in script files.
- Boundary markers: None identified in the instructions for isolating paper content from the agent's internal instructions.
- Capability inventory: Execution of system commands for file conversion and video assembly, and network requests for TTS API interactions.
- Sanitization: No validation or sanitization of input text before processing or interpolation into prompts.
Audit Metadata