beautiful-mermaid

Warn

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The render.ts script is configured to dynamically download and install the beautiful-mermaid package from public registries (NPM or Bun) if it is not already present in the environment.
  • [REMOTE_CODE_EXECUTION]: By automatically fetching and installing external dependencies at runtime, the skill potentially executes unverified code from remote sources.
  • [COMMAND_EXECUTION]: The script scripts/render.ts uses child_process.execSync to invoke system package managers (npm install or bun add) to handle missing dependencies.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text descriptions to generate Mermaid diagram code, creating a surface for embedded malicious instructions.
  • Ingestion points: User descriptions or provided Mermaid code processed in SKILL.md (Step 1).
  • Boundary markers: None identified in the prompt generation or rendering workflow.
  • Capability inventory: File system access (node:fs), subprocess execution for package installation (execSync).
  • Sanitization: No input validation or sanitization is performed on the Mermaid code before rendering or embedding it in HTML.
  • [DYNAMIC_EXECUTION]: The create-html.ts script constructs an HTML file by directly embedding the raw content of an SVG file. If the SVG (generated from user-controlled input) contains malicious <script> tags, they will execute when the file is opened by the agent-browser tool, potentially leading to local file access or data exfiltration from the browser context.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 14, 2026, 08:42 AM