three-radius-model

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 is an adaptive coaching dialogue that elicits user-provided case details (e.g., “Tell me about a specific domain…” and subsequent gated steps), meaning outsider-authored free text is directly ingested from the user into the LLM context.