backlinks-timeseries-new-lost-summary

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow calls an external Frevana API at runtime and then reads the returned HTTP response body as text (curl -o "$RESPONSE_FILE" followed by raw = response_path.read_text(...) and cat "$RESULT_FILE"), so any outsider-authored free text in that response (even if JSON) is ingested into the agent’s LLM context via stdout.