backlinks-list

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

• Third-party content exposure detected (high risk: 0.75). Yes—this skill calls a runtime HTTP endpoint and then reads the returned response body as text (raw = response_path.read_text(...)) and prints it to stdout (cat "$RESULT_FILE"), so any free-form JSON content from the outsider (Frevana API response) is ingested into the agent’s LLM context.