media-to-roadmap

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow explicitly fetches and ingests outsider-authored free text (e.g., YouTube/video transcripts, blog articles/docs pages, and login-gated X/Twitter thread text via search/mirrors, then indexes/queries it for concrete techniques) as the basis for the roadmap.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches external source content at runtime (e.g., it instructs fetching raw files from raw.githubusercontent.com) and uses that fetched content to drive and control the model's outputs (roadmap generation), so raw.githubusercontent.com is flagged.