ducksearch

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly exposes arbitrary public web content via commands like "ducksearch fetch https://example.com" and tools "DuckDuckGoWebSearch" and "UrlContentExtractor", meaning the agent will ingest and act on untrusted third-party webpages and search results that could contain injected instructions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes "npx -y ducksearch" at runtime (which fetches and executes the ducksearch package from the npm registry, e.g. https://registry.npmjs.org/ducksearch), so remote code is fetched-and-run and is required for the skill to operate.