runtime-observer

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

• Third-party content exposure detected (high risk: 0.80). The skill directly fetches user-generated telemetry (logs, traces, exceptions) from a Runtime Observer collector (endpoints such as /v1/agent/logs, /v1/agent/traces/{id} and /v1/agent/traces/{id}/context) and the SKILL.md and client/CLI explicitly instruct using trace-context markdown as LLM input, so untrusted log/trace content could inject instructions that alter agent decisions or tool use.