langfuse-observability

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

• Third-party content exposure detected (high risk: 0.70). The skill’s required workflow feeds outsider-authored free text into the agent LLM context via updateActiveObservation({ input: messages }) where messages are runtime chat inputs (which can include user/third-party text) and are sent to Langfuse instrumentation/output (output: response.choices[0].message.content), so any such text can be ingested into the LLM-relevant tracing context.