test-research-code

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow reads outsider-authored free text from the target research-code directory at runtime (e.g., top-level README and other source files via repro_check.py using path.read_text(...)), which can contain prompt-injection content that is then included in the agent’s LLM context through the generated audit report.