wearables-ui-toolkit-web

Pass

Audited by Gen Agent Trust Hub on Nov 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes scripts/init-app.mjs, which performs file system operations to set up a new application workspace. It also instructs the agent to run common development tools such as vite and tsc, as well as specific validation scripts located in sibling directories.\n- [EXTERNAL_DOWNLOADS]: The assets/app-shell/package.json file specifies several dependencies from the @wearables-ui-toolkit namespace. These are official packages associated with the Meta Ray-Ban Display toolkit and are necessary for the application's runtime and build process.\n- [INDIRECT_PROMPT_INJECTION]: The scripts/init-app.mjs script accepts a user-provided workspace path as a command-line argument, which is used to create directories and copy template files.\n
  • Ingestion points: The script reads the workspace path from process.argv[2].\n
  • Boundary markers: SKILL.md defines a "Hard execution boundary" that explicitly restricts the agent's file system access and tool usage during the development process.\n
  • Capability inventory: The script uses mkdirSync for directory creation and copyFileSync for transferring static template files (index.html, package.json, tsconfig.json, vite.config.ts) to the target workspace.\n
  • Sanitization: The input path is processed using path.resolve but lacks further validation against arbitrary path creation; however, the risk is mitigated by the fact that only fixed internal assets are written.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 3, 2026, 06:17 PM