claude-real-video-for-agents

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md’s required agent workflow runs crv "<url-or-path>" ... and then reads crv-out/MANIFEST.txt (which “includes the transcript”) and other generated outputs; when the source is a runtime-supplied URL/video from someone else, the transcript/on-screen text is outsider-authored free text ingested into the LLM context via those output files.