next-upgrade

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md:9-14 instructs using WebFetch to retrieve official Next.js upgrade documentation from nextjs.org at runtime, which is public web content (outsider-authored free text) that could be ingested into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs using WebFetch at runtime to retrieve and follow official Next.js upgrade documentation (e.g., https://nextjs.org/docs/app/guides/upgrading/codemods and https://nextjs.org/docs/app/guides/upgrading/version-16), so the fetched remote content would directly control the agent's upgrade instructions.