The Agent Skills Directory

[Command Injection Prevention]: The guide explicitly recommends quoting interpolated paths when building shell commands for sandbox.run or sandbox.spawn. This is a documented best practice to mitigate potential command injection risks when handling file paths that might contain spaces or special characters.
[Secure Credential Management]: The instructions direct developers to use a centralized authentication helper (getAiGatewayAuthFromEnv) instead of reading environment variables directly. This encourages a standardized and more secure approach to resolving credentials for AI gateways.
[Trusted Infrastructure Integration]: The skill provides guidance for integrating with first-party infrastructure and references official Vercel repositories. The use of standard package managers and build tools aligns with expected developer workflows.

add-harness-package