AI军事情报定位与常规武器能力评估方法论：kill chain各环节的模型能力基准——身份关联/分类(模拟社交媒体200任务三难度, F1)、照片地理定位(YFCC100M, 前沿模型中位误差37km超人类最强GeoGuessr选手151km)、文本地理定位(GeoText, 沙箱搜索工具)、无人机打击精度等。发现前沿模型定位能力接近超人类，开放权重模型(Kimi K3)落后一代但仍有令人担忧的能力。适用于AI滥用风险评估、军事情报能力基准设计。Activation: intelligence targeting evaluation, geolocation benchmark, kill chain AI, identity correlation, AI weapons capability, surveillance risk, model misuse assessment
no risky patterns detected
Type: Agent skill (SKILL.md) · Category: Maps, Travel & Location · Source: https://github.com/hiyenwong/ai_collection/tree/HEAD/collection/skills/ai-intell
| Capabilities (fewer powerful permissions = higher) | 50/50 |
| Maintenance | 21/35 |
| Documentation | 15/15 |
Always review files before installing. How scores work. Request a correction