live-demo
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions include cloning an external repository from GitHub (
https://github.com/new-xp/ultrademo) to the local environment. - [REMOTE_CODE_EXECUTION]: The skill executes code from the downloaded external repository using
npm runcommands (npm run doctor,npm run capture,npm run tts,npm run render) after installing dependencies withnpm ci. - [INDIRECT_PROMPT_INJECTION]: The skill browses and captures content from live applications, which introduces a risk of indirect prompt injection from the target app's content.
- Ingestion points: UI elements and page content from target applications accessed via Playwright.
- Boundary markers: The skill contains an explicit instruction to ignore and flag any fetched content that appears to address the agent directly.
- Capability inventory: Shell command execution via
npm run, browser automation, and local file system access for media storage. - Sanitization: Includes instructions for manual redaction of sensitive user data and sweeping final frames for visible secrets.
Audit Metadata