Summarize

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

• Third-party content exposure detected (high risk: 0.85). The required workflow ingests outsider-authored free text by fetching and summarizing runtime web content from user-supplied URLs (public pages/articles) and/or extracting YouTube transcripts from third-party videos, which the LLM then reads into its context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill fetches arbitrary remote URLs at runtime (e.g., https://example.com and https://youtu.be/dQw4w9WgXcQ) and injects that fetched page/transcript content into the model context for summarization, so external content can directly control the agent's prompts.