presentation
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes several shell commands to manage the presentation project lifecycle. This includes scaffolding new Slidev projects using
npm create slidev, installing dependencies withnpm install, and running local development servers viapnpm dev. - [EXTERNAL_DOWNLOADS]: The skill initiates downloads from external sources to set up its environment. It installs the official Slidev agent skill using
npx skills add slidevjs/slidev, fetches Node.js packages such asshikiandmermaid, and downloads the Chromium browser binary vianpx playwright install chromiumfor rendering purposes. These downloads originate from official and well-known service providers. - [INDIRECT_PROMPT_INJECTION]: The skill features an extensive research and grounding workflow documented in
references/research-grounding.md. It is designed to ingest data from web searches, GitHub repositories, and documentation sites to build the presentation brief. This ingestion of untrusted external content creates a surface for indirect prompt injection, as the skill lacks explicit sanitization or instruction-isolation markers for the processed data. - [DYNAMIC_EXECUTION]: To produce high-quality visual assets and verify slide rendering, the skill uses Playwright to run a headless browser. As seen in
assets/render-asset.mjsandreferences/verification.md, it renders HTML, CSS, and Mermaid diagrams and takes screenshots of the slides. This process involves the dynamic execution of code within the browser context based on the content of the slides.
Audit Metadata