perplexity-architecture-variants

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime workflow includes accepting outsider-provided free text via an HTTP endpoint/body (e.g., req.body.query/request.json["query"]) and sending it directly as the messages[].content prompt to Perplexity/LLM (/api/search in the Direct Widget variant, and similarly in the cached/pipeline variants via search(query) and researchTopic(topic)).