eval-otel

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow runs ./bin/agent.sh ... --otel and then reads /tmp/agent_eval_otel_smoke/rows.jsonl plus raw_ref log files; those artifacts are produced by the eval harness and the tested runtimes (e.g., Codex/Claude/Louie) and can include free-form text originating from outside the operating user (model outputs and/or runtime-generated prompt/trace content), which the agent then ingests for inspection/reporting.