send-experiment-designer

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md, the skill’s runtime workflow can read untrusted free text contained in a “finished ESP results export” supplied by the user (and other manually exported own-data fields), meaning outsider-authored text can be injected through the user-provided results export payload that the LLM ingests at runtime.