model-pruning

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

• Third-party content exposure detected (high risk: 0.85). The required runtime workflow ingests calibration text from an external dataset source (e.g., load_dataset("wikitext", ...) / load_dataset("allenai/c4", ...)), which is outsider-authored free-form web content that is tokenized into LLM-readable text for forward passes during pruning.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill's installation and deployment steps explicitly clone and run code from external GitHub repositories (e.g., git clone https://github.com/locuslab/wanda and git clone https://github.com/IST-DASLab/sparsegpt followed by pip install / python main.py), meaning these URLs fetch and execute remote code that the skill relies on at runtime.