content-strategy

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

• Third-party content exposure detected (high risk: 0.95). At runtime the workflow calls UnifAPI to fetch public web/community text (e.g., seo/serp for site:reddit.com <topic> and then reddit/posts/{id}/comments), which ingests outsider-authored free-form Reddit comments into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill explicitly requires runtime calls to the UnifAPI MCP endpoint (e.g., https://unifapi.com/mcp) via the unifapi skill to fetch live SEO/Reddit/YouTube/News/TikTok content which is injected into the agent context and directly drives the strategy prompts/outputs.