pptx-deck-creation

Warn

Audited by Gen Agent Trust Hub on Jul 25, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on the agent writing and executing local Python scripts to interact with the PowerPoint object model and the file system, specifically for building decks and analyzing OOXML structures.
  • [REMOTE_CODE_EXECUTION]: The skill includes Python code templates for the agent to execute at runtime that perform network requests using urllib.request and download assets from external servers.
  • [EXTERNAL_DOWNLOADS]: The agent is instructed to fetch design system data from getdesign.md and icons from api.iconify.design to inform deck creation.
  • [PROMPT_INJECTION]: Processing external PowerPoint files and live web content from getdesign.md exposes the agent to indirect prompt injection. The skill includes a defense-in-depth instruction to ignore embedded commands in these sources, which mitigates but does not eliminate the risk.
  • [DATA_EXFILTRATION]: The skill includes functionality to send prompts and potentially source material to external AI providers (OpenAI/Azure OpenAI) to generate infographics, though it explicitly requires user confirmation before doing so.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 25, 2026, 12:43 PM
Security Audit — agent-trust-hub — pptx-deck-creation