create-write-like-me

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md, steps 2 and 7 (“Gather evidence” and “Forward-test blindly”) explicitly ingest and use user-provided evidence and raw test inputs/outputs during the calibration and isolated matrix execution, making the workflow consume outsider-authored free text at runtime.