pm-preview
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes document filenames and content from the docs/ directory and renders them in a web browser, creating a potential surface for cross-site scripting (XSS).
- Ingestion points: File names and document content are read from the local docs/ directory via preview-server.cjs.
- Boundary markers: None identified in the prompt interpolation or rendering process.
- Capability inventory: The server injects a document tree directly into a script block, and the browser-side helper script renders Markdown content into the DOM.
- Sanitization: Filenames are injected into the HTML template using JSON.stringify without escaping HTML closing tags (e.g., ), and Markdown content is rendered using innerHTML without explicit sanitization.
- [EXTERNAL_DOWNLOADS]: The preview interface fetches a Markdown rendering library from a well-known content delivery network.
- Fetches marked.js from https://cdn.jsdelivr.net/npm/marked/marked.min.js.
Audit Metadata