brainrotshorts

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

• Third-party content exposure detected (high risk: 0.75). The required workflow includes the Reddit URL path (reddit_video_create_from_url / reddit_post_fetch) which fetches outsider-authored Reddit post text at runtime and then uses it to generate/adapt scripts, feeding that free text into the agent’s LLM context via tool results.

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

• Potentially malicious external URL detected (high risk: 0.80). The agent configuration includes a runtime MCP tool endpoint (https://www.brainrotshorts.com/api/mcp) in agents/openai.yaml which the agent will call to obtain tool behavior/schemas that can directly influence agent instructions at runtime.