From market-radar
This skill should be used when the intelligence-cluster agent needs guidance on clustering intelligence cards to themes, detecting new themes, and determining card-theme assignments.
How this skill is triggered — by the user, by Claude, or both
Slash command
/market-radar:clustering-methodologyThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
情报聚类方法论,指导如何将情报卡片分配到合适的主题,以及如何检测新主题。
情报聚类方法论,指导如何将情报卡片分配到合适的主题,以及如何检测新主题。
阅读情报卡片的以下部分:
title, primary_domain, secondary_domains, intelligence_date, 领域特定字段## 核心事实 章节对每张卡片,按以下规则匹配主题:
检查卡片的 primary_domain 和 secondary_domains 是否在主题的 domains 列表中:
如果 primary_domain ∈ theme.domains → 基础分 +0.3
如果 secondary_domains ∩ theme.domains ≠ ∅ → 附加分 +0.1
检查卡片标题和正文是否包含主题关键词:
对于 theme.keywords 中的每个关键词:
如果出现在标题 → 分数 +0.2
如果出现在正文 → 分数 +0.1
检查卡片中的实体是否匹配主题的 match_rules.entity_patterns:
如果卡片提及 entity_patterns 中的实体 → 分数 +0.15
confidence = min(1.0, 基础分 + 关键词分 + 实体分)
根据置信度决定卡片归属:
| 置信度范围 | 决策 |
|---|---|
| >= 0.7 | 明确分配到该主题 |
| 0.4 - 0.7 | 边缘情况,标记为候选主题 |
| < 0.4 | 不分配到该主题 |
多主题情况:
当以下情况发生时,考虑推荐新主题:
1. 存在 >= min_cards 张未分配卡片(置信度 < 0.4)
2. 这些卡片有相似的特征:
- 相同或相近的 primary_domain
- 共同的关键词或实体
- 相似的内容主题
分析未分配卡片的共同特征:
1. 统计 primary_domain 分布
2. 提取高频关键词(出现 >= 3 次)
3. 识别共同实体
4. 归纳内容主题
{
"suggested_theme_id": "suggested-theme-slug",
"name": "建议的主题名称",
"description": "主题描述",
"keywords": ["关键词1", "关键词2"],
"domains": ["相关领域"],
"supporting_cards": ["card1.md", "card2.md"],
"confidence": 0.85
}
聚类时重点关注:
threat_actor)threat_type)常见主题:
聚类时重点关注:
tech_name)maturity)常见主题:
聚类时重点关注:
vendor_name)business_area)常见主题:
当卡片同时匹配多个主题时:
1. 列出所有候选主题及置信度
2. 如果都是强匹配(>= 0.7),分配到多个主题
3. 如果置信度相近,优先选择 primary_domain 匹配的主题
当卡片主题匹配但内容不符时:
1. 降低置信度
2. 记录不匹配原因
3. 考虑是否需要调整主题配置
当发现新的情报模式时:
1. 收集相似卡片
2. 分析共同特征
3. 推荐新主题给用户确认
聚类结果返回 JSON:
{
"clustered": [
{
"card_path": "Threat-Landscape/20260301-lockbit-ransomware.md",
"assigned_themes": ["ransomware-threats"],
"confidence": {"ransomware-threats": 0.85}
}
],
"edge_cases": [
{
"card_path": "Emerging-Tech/20260301-ai-cloud-security.md",
"candidate_themes": ["ai-security", "cloud-security"],
"confidence": {"ai-security": 0.55, "cloud-security": 0.52},
"reason": "内容涉及两个主题,置信度相近"
}
],
"unclustered": [
{
"card_path": "Industry-Analysis/20260301-iot-security-market.md",
"reason": "无匹配主题"
}
],
"suggested_new_themes": [
{
"suggested_theme_id": "iot-security",
"name": "IoT 安全",
"description": "物联网安全市场、技术和威胁动态",
"keywords": ["IoT安全", "物联网安全", "OT安全"],
"domains": ["Emerging-Tech", "Industry-Analysis"],
"supporting_cards": ["..."],
"confidence": 0.8
}
]
}
聚类完成后执行:
references/domain-guidance.md - 各情报领域的聚类特定指导npx claudepluginhub cyberstrat-forge/cyber-nexus --plugin market-radarGroups relevance-filtered papers into 2-5 argument-shaped thematic clusters using abstract semantics and keyword overlap, naming clusters by what they argue rather than their topic. Surfaces an outliers bucket and assigns rationale.
Organizes qualitative research data from interviews, observations, or surveys into affinity diagrams with clusters, themes, insight statements, and prioritization.
Manages end-to-end cyber threat intelligence lifecycle from planning through collection, analysis, dissemination, and feedback. Helps establish CTI programs, define intelligence requirements, and build feedback loops. Activates on CTI program maturity, intelligence requirements, PIRs.