linkfox-aigc-imagegen-product

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

• Third-party content exposure detected (medium risk: 0.30). 运行时在套图的 plan 阶段会把用户输入/上下文（尤其用户的卖点 point、画面描述 imageDesc、以及历史卖点等）拼入用于 textgen 的提示词模板并调用 textgen，从而让先于选定具体模板项的“任意用户文本”进入 LLM 处理链路（S1-content-reasoning → scripts/collection_pipeline/plan.py/_build_s1_prompt + run_textgen，随后生成 imagePlanList 与下游出图 prompts）。

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

• Potentially malicious external URL detected (high risk: 0.90). The SKILL.md (runtime) explicitly instructs downloading and installing https://agent-files.linkfox.com/skills/linkfox-onboarding/release.zip when the onboarding skill is missing, which fetches remote skill code at runtime that could execute or change agent behavior.