ablate-ai-layer

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

• Third-party content exposure detected (medium risk: 0.30). The runtime path run_ablation.py reads outsider-authored free text into the prompt via prompt = Path(args.task_file).read_text(...).strip() and then passes it to the agent on stdin (subprocess.run(..., input=prompt, ...)).