ppt
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
playwright-coreto launch and control headless browser processes (Chromium, Chrome, or Edge) on the local system for slide rendering and capture. This is a standard part of its automated design and export workflow.\n- [EXTERNAL_DOWNLOADS]: During slide rendering, templates may load the Chart.js library from the JSDelivr CDN. Additionally, the skill setup may involve downloading browser binaries via Playwright's installation command.\n- [PROMPT_INJECTION]: The skill implements an indirect prompt injection surface by ingesting external web content via a research workflow, as described inreferences/research.md.\n - Ingestion points: External data enters the agent context through research logs archived in
.pptwork/<deck>/materials/research.md.\n - Boundary markers: The skill lacks explicit isolation markers to wrap external content, relying instead on markdown structure.\n
- Capability inventory: The skill has capabilities for filesystem manipulation (
deck.mjs), script execution for browser rendering (screenshot.mjs), and file export (export.mjs).\n - Sanitization: Content is summarized by the agent without specific programmatic sanitization of the input data.
Audit Metadata