The Agent Skills Directory

[COMMAND_EXECUTION]: The assets/workflows/deploy.yml workflow is vulnerable to shell command injection in the 'Deploy to Cloudflare Workers' step. The inputs.message parameter is directly interpolated into a shell variable assignment: msg="${{ inputs.message }}". An attacker who controls this input (e.g., via workflow_dispatch) can execute arbitrary commands on the runner by including shell metacharacters like double quotes and semicolons.\n- [REMOTE_CODE_EXECUTION]: In assets/workflows/deploy.yml, the MoonBit toolchain is installed using curl -fsSL https://cli.moonbitlang.com/install/unix.sh | bash. Piped remote execution from external domains is a high-risk pattern as it permits arbitrary code execution if the source script or domain is compromised.\n- [EXTERNAL_DOWNLOADS]: The deployment process uses npx -y @dotenvx/dotenvx@latest to manage secrets. Pulling the latest version of a package from the npm registry at runtime without version pinning introduces a supply chain risk where a malicious package update could compromise the deployment environment.\n- [EXTERNAL_DOWNLOADS]: The workflow utilizes standard GitHub Actions such as actions/checkout@v4 and actions/setup-node@v4 to prepare the environment.

cloudflare-workers-cd-rollback