vercel-skill-cli-expert

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 1.00). The CLI explicitly installs skills from arbitrary public repositories (e.g., "npx skills add owner/repo" and full GitHub/GitLab/any git URL in references/commands.md) and the Skill Authoring Guide and progressive loading docs state agents load and execute SKILL.md instructions (agent reads the SKILL.md body and referenced scripts), so untrusted third-party, user-generated content can directly influence agent actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The CLI explicitly installs skills from remote git/GitHub URLs (e.g., https://github.com/vercel-labs/agent-skills), which at runtime fetches SKILL.md and related repo files that directly define agent instructions/prompts and can be executed or loaded into agent context.