qoderwork-ppt
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
- [UNVERIFIABLE_DEPENDENCIES]: The skill instructions direct the agent to run
npm installfor dependencies and the code fetches images from remote URLs to convert them into data URLs for inclusion in slides. - [DYNAMIC_EXECUTION]: The tool uses Puppeteer to render generated HTML templates containing user data before converting the rendered output into a PPTX file.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user requirements and interpolates them into HTML templates that are rendered in a browser environment, presenting an attack surface for indirect injection.
- Ingestion points: User-provided topic and requirements in SKILL.md (Step 1).
- Boundary markers: None identified in the HTML generation logic.
- Capability inventory: Shell command execution (
node), network requests (fetch), and browser execution (puppeteer). - Sanitization: Content is largely treated as text, but
richtextfields useinnerHTMLwithout sanitization.
Audit Metadata