long-context

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

• Third-party content exposure detected (high risk: 0.90). This skill's SKILL.md explicitly instructs fetching and ingesting public third‑party content (e.g., git clone https://github.com/jquesnelle/yarn, HuggingFace model loads like from_pretrained("meta-llama/..."), and datasets via load_dataset("pg19","scientific_papers","codeparrot/github-code")), so the agent would read and act on untrusted web-hosted code/data that could contain instructions influencing tool use.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs cloning and running code from the YaRN repository (git clone https://github.com/jquesnelle/yarn and running python scripts/train.py), which fetches remote code and executes it at runtime, making that URL a required dependency that runs remote code.