vercel-deploy

Fail

Audited by Gen Agent Trust Hub on Mar 25, 2026

Risk Level: HIGHDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The script scripts/deploy.sh packages the user's project directory into a tarball and uploads it to https://codex-deploy-skills.vercel.sh/api/deploy. While this domain is hosted on Vercel, it is an unofficial intermediary endpoint controlled by the skill author, exposing potentially sensitive source code to a third party.
  • [COMMAND_EXECUTION]: The skill uses scripts/deploy.sh to perform file system operations, package data using tar, and execute network requests via curl. It also expects the vercel CLI to be available for direct deployment commands.
  • [PRIVILEGE_ESCALATION]: The SKILL.md file explicitly instructs the agent to use sandbox_permissions=require_escalated if network access is restricted. This encourages the agent to bypass standard sandbox security constraints to perform its functions.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Mar 25, 2026, 01:22 PM
Security Audit — agent-trust-hub — vercel-deploy