qwen-image-edit

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). These links include third‑party GitHub repositories and direct Hugging Face model file downloads (large .safetensors binaries) hosted by individual authors rather than official vendors, which are common vectors for unvetted code or payload distribution and therefore warrant caution.

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

• Third-party content exposure detected (medium risk: 0.30). The runtime path references/runtime.md → scripts/prepare_workflow.py ingests outsider-provided free text as the edit prompt (written to prompt.txt and patched into assets/workflows/qwen-image-edit.api.json / qwen-image-edit.ui.json), so the agent can be influenced directly by user text without selecting a specific item first.