baoyu-imagine

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

• Third-party content exposure detected (high risk: 0.90). The skill accepts and forwards remote http(s) reference-image URLs (e.g., "http(s):// URLs are forwarded as‑is" in references/providers/dashscope.md and the batch/ref handling in scripts/main.ts — resolveBatchReferencePath / createTaskArgs / loadBatchTasks), which are untrusted third‑party inputs that get consumed by provider models during generation and can therefore indirectly inject instructions into the generation workflow.