create-canvas-app

Pass

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The toolkit includes a script scripts/vendor-lucide.mjs that uses the system package manager to install the lucide-react package into a temporary directory for icon data extraction. The version string provided to the script is strictly validated against a semantic versioning regular expression to prevent command injection.
  • [EXTERNAL_DOWNLOADS]: Development and testing dependencies include @microsoft/vally, which are well-known packages from a established technology provider.
  • [COMMAND_EXECUTION]: The toolkit provides several utility scripts (scripts/new-canvas.mjs, scripts/sync-kit.mjs) and a PowerShell installer (scripts/install-local.ps1) that perform standard file system operations such as mirroring directories, creating folders, and writing boilerplate code files to scaffold new extensions.
  • [SAFE]: The skill includes a dedicated networking module kit/net.mjs that implements a server-side SSRF guard. This guard validates hostnames and IP addresses to prevent the agent from accessing loopback, private, or cloud metadata ranges.
  • [SAFE]: The kit implements a dependency-free JSON schema validator in kit/validate.mjs that is used to enforce input and state contracts. The validator explicitly uses Object.hasOwn checks to prevent prototype pollution and bypass attempts using inherited property names.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 7, 2026, 01:58 PM
Security Audit — agent-trust-hub — create-canvas-app