together-images

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly accepts and fetches arbitrary public URLs for reference images and LoRA adapters (see references/api-reference.md "Reference Images" and "LoRA Adapters" sections), and the example scripts (scripts/kontext_editing.py with download_image, scripts/lora_generation.py) plus SKILL.md workflow step 3 show the agent will retrieve and use those external, untrusted resources to influence generation/editing, creating a clear vector for indirect prompt injection.