unsloth-sft

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly ingests and reformats public, user-generated instruction datasets such as Alpaca and ShareGPT as part of its workflow (see "When to Use", the conversation_extension workflow, and the scripts like scripts/unsloth-sft_tool.js/.py that convert Alpaca/ShareGPT items), so it will read untrusted third‑party content that could carry indirect prompt injections.