vercel-deploy

SUSPICIOUS. The stated purpose fits deployment, and the visible domains are official Vercel endpoints, so the scanner's free-hosting flags are weak. The main concern is that the skill's actual behavior is delegated to an unseen shell script while also claiming unauthenticated deployment, which is inconsistent with normal Vercel API auth expectations and prevents verification of data flow, packaging scope, and credential handling.

No clear malware/backdoor behavior is evident in the shown snippet. The main security risk is supply-chain-style data egress: the script packages nearly all contents of the provided project directory (excluding only node_modules and .git) and uploads the resulting archive to a remote DEPLOY_ENDPOINT via HTTP POST. If DEPLOY_ENDPOINT is untrusted/misconfigured or the project contains sensitive files not excluded here, this can lead to data exfiltration or unintended distribution. Additionally, raw server responses are echoed and JSON fields are parsed via regex, creating log/data exposure and fragility risks rather than direct malicious code execution.