vercel-deploy

Warn

Audited by Socket on Feb 28, 2026

1 alert found:

Security
SecurityMEDIUM
SKILL.md

This skill implements an anonymous deploy helper that packages and uploads a project to produce preview/claim URLs. It is not overtly malicious in the provided manifest, but it presents moderate to high security risks: accidental exfiltration of secrets (due to minimal exclusion list), execution of local script code with broad filesystem access, lack of endpoint transparency, and guidance that encourages widening network egress. Recommendations: require/implement explicit exclusion of sensitive file patterns (.env, .netrc, *key, *.pem, secrets), prompt users to confirm inclusion of non-source files before upload, document exact upload endpoints and TLS/auth handling, avoid instructing users to add wildcard domains (prefer explicit hostnames), and avoid executing unreviewed scripts or run them in a strict sandbox. With mitigations, the functional behavior can be acceptable for general deployments; without them, use in sensitive environments is risky.

Confidence: 98%Severity: 75%
Audit Metadata
Analyzed At
Feb 28, 2026, 08:36 PM
Package URL
pkg:socket/skills-sh/pm-minji%2Fagent-skills%2Fvercel-deploy%2F@5558e8633e47a6ba3a2f35cebf974bf1aea115c3