hardware-workload-estimator

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

• Third-party content exposure detected (high risk: 0.75). High risk of outsider-authored free text: Step 4 in SKILL.md uploads/executes an auto-generated Colab notebook that embeds script_content read from a provided training script path (scripts/generate_colab_notebook.py reads arbitrary file text via open(args.script, ...), then writes it into .ipynb source cells that are later uploaded to https://colab.research.google.com/ and executed).