Skip to content

Hypergraph is more powerful. Transform unstructured text into structured knowledge with LLMs. Graphs, hypergraphs, and spatio-temporal extractions — with one command.

docs: 更新README文档并添加多个示例模板

yifanfeng97c4916cbd9d9e
diff --git a/README.md b/README.md
index 92dfb17..737301b 100644
--- a/README.md
+++ b/README.md
@@ -96,18 +96,15 @@ Transform LLM output from scattered text into **searchable, queryable, and reaso
---
-## 🌍 12 Domains, 200+ Ready-to-Use Templates
+## 🌍 6 Domains, 40+ Ready-to-Use Templates
> Pre-built templates for **different industries** — extract knowledge **right away** without designing schemas.
| Domain | Templates | Domain | Templates |
| :--- | :---: | :--- | :---: |
-| 💰 Finance | 25+ | 📜 History | 12+ |
-| 🏥 Medicine | 20+ | 🧬 Biology | 10+ |
-| ⚖️ Legal | 15+ | 🎭 Literature | 10+ |
-| 🌿 TCM | 15+ | 📰 News | 12+ |
-| 🔧 Industry | 18+ | 🌾 Agriculture | 8+ |
-| 🍜 Food | 8+ | 🌐 General | 20+ |
+| 📚 General | 16 | 💰 Finance | 21 |
+| 🏥 Medicine | 15 | 🌿 TCM | 10 |
+| ⚖️ Legal | 11 | 🔧 Industry | 15 |
**✅ Bilingual** — Templates available in both English and Chinese
diff --git a/README_ZH.md b/README_ZH.md
index 2397711..d7648dd 100644
--- a/README_ZH.md
+++ b/README_ZH.md
@@ -96,18 +96,15 @@ he show kb # 可视化
---
-## 🌍 12 个领域,200+ 开箱即用模板
+## 🌍 6 个领域,40+ 开箱即用模板
> 针对**不同行业**预置的模板 —— 无需设计schema,直接提取知识。
| 领域 | 模板数 | 领域 | 模板数 |
| :--- | :---: | :--- | :---: |
-| 💰 金融 | 25+ | 📜 历史 | 12+ |
-| 🏥 医疗 | 20+ | 🧬 生物 | 10+ |
-| ⚖️ 法律 | 15+ | 🎭 文学 | 10+ |
-| 🌿 中医 | 15+ | 📰 新闻 | 12+ |
-| 🔧 工业 | 18+ | 🌾 农业 | 8+ |
-| 🍜 美食 | 8+ | 🌐 通用 | 20+ |
+| � 通用 | 16 | � 金融 | 21 |
+| 🏥 医疗 | 15 | 🌿 中医 | 10 |
+| ⚖️ 法律 | 11 | 🔧 工业 | 15 |
**✅ 双语支持** — 模板同时支持中文和英文
diff --git a/hyperextract-skills/README.md b/hyperextract-skills/README.md
new file mode 100644
index 0000000..be2ccf0
--- /dev/null
+++ b/hyperextract-skills/README.md
@@ -0,0 +1,108 @@
+# Hyper-Extract Skills
+
+Knowledge template design skills for Hyper-Extract.
+
+## Skills Overview
+
+| Skill | Purpose |
+|-------|---------|
+| Root | Entry point for template design |
+| brainstorm | Requirements exploration and type discussion |
+| record-designer | Design model/list/set structures |
+| graph-designer | Design graph/hypergraph/etc structures |
+| yaml-validator | Validate YAML configurations |
+| multilingual | Convert to multi-language support |
+
+## Installation
+
+### Claude Code
+
+Copy the `hyperextract-skills` folder to your Claude Code skills directory:
+
+```bash
+cp -r hyperextract-skills ~/.claude/skills/
+```
+
+Or use the plugin command:
+
+```bash
+/plugin install hyperextract-skills
+```
+
+### Trae
+
+Copy the `hyperextract-skills` folder to your Trae skills directory:
+
+```bash
+cp -r hyperextract-skills ~/.trae/skills/
+```
+
+## Usage
+
+### Quick Start
+
+1. Start with the brainstorm skill to explore your requirements
+2. Based on the recommended type, use either record-designer or graph-designer
+3. Optionally validate with yaml-validator
+4. Optionally convert to multilingual
+
+### Example Workflow
+
+```
+You: I want to extract key information from financial reports
+
+Assistant (using brainstorm):
+Let's explore your requirements...
+
+You: I need to extract company name, revenue, and quarter from earnings calls
+
+Assistant (using brainstorm):
+Based on your description, I recommend: model type
+
+Assistant (using record-designer):
+Let's design the fields for your model...
+```
+
+## Supported Types
+
+### Record Types
+- **model**: Single structured object
+- **list**: List of homogeneous objects
+- **set**: Deduplicated object set
+
+### Graph Types
+- **graph**: Binary relation graph
+- **hypergraph**: Multi-entity relation hypergraph
+- **temporal_graph**: Temporal graph with time dimension
+- **spatial_graph**: Spatial graph with location dimension
+- **spatio_temporal_graph**: Combined temporal and spatial
+
+## Directory Structure
+
+```
+hyperextract-skills/
+├── SKILL.md # Root skill
+├── brainstorm/
+│ └── SKILL.md # Requirements exploration
+├── record-designer/
+│ └── SKILL.md # Record type design
+├── graph-designer/
+│ └── SKILL.md # Graph type design
+├── yaml-validator/
+│ └── SKILL.md # YAML validation
+├── multilingual/
+│ └── SKILL.md # Multi-language conversion
+└── examples/ # Example templates
+ ├── model/
+ ├── list/
+ ├── set/
+ ├── graph/
+ ├── hypergraph/
+ ├── temporal_graph/
+ ├── spatial_graph/
+ └── spatio_temporal_graph/
+```
+
+## License
+
+Apache 2.0
diff --git a/hyperextract-skills/SKILL.md b/hyperextract-skills/SKILL.md
new file mode 100644
index 0000000..f4d9f49
--- /dev/null
+++ b/hyperextract-skills/SKILL.md
@@ -0,0 +1,69 @@
+---
+name: hyper-extract
+description: |
+ Hyper-Extract Knowledge Template Design Assistant. Helps users design YAML configuration templates through natural language dialogue.
+
+ ## Supported Types (8 AutoTypes)
+ ### Record Types
+ - model: Single structured object
+ - list: List of homogeneous objects
+ - set: Deduplicated object set
+ ### Graph Types
+ - graph: Binary relation graph
+ - hypergraph: Multi-entity relation hypergraph
+ - temporal_graph: Temporal graph with time dimension
+ - spatial_graph: Spatial graph with location dimension
+ - spatio_temporal_graph: Combined temporal and spatial
+
+ ## Workflow
+ 1. brainstorm: Explore requirements, discuss type details
+ 2. record-designer OR graph-designer: Design structure + rules
+ 3. yaml-validator: Validate configuration (optional)
+ 4. multilingual: Multi-language conversion (optional)
+---
+
+# Hyper-Extract Knowledge Template Design Assistant
+
+## How to Use
+
+### Mode 1: Start from Requirements
+Describe your scenario, we'll explore together to clarify needs and determine the type.
+
+### Mode 2: Start from Type
+Already know the type? Jump directly to the designer.
+
+### Mode 3: Start from Example
+Have a reference template? We'll analyze and adapt it.
+
+## Sub-Skills
+
+| Skill | Purpose |
+|-------|---------|
+| brainstorm | Deep discussion about requirements and type-specific details |
+| record-designer | Design model/list/set structures |
+| graph-designer | Design graph/hypergraph/etc structures |
+| yaml-validator | Validate YAML correctness |
+| multilingual | Multi-language support |
+
+## Calling Sub-Skills
+
+```
+[Invoke brainstorm]
+User description: [description]
+Current understanding: [if any]
+
+[Invoke record-designer]
+Type: [model/list/set]
+Requirements: [from brainstorm]
+
+[Invoke graph-designer]
+Type: [graph/hypergraph/temporal_graph/spatial_graph/spatio_temporal_graph]
+Requirements: [from brainstorm]
+
+[Invoke yaml-validator]
+YAML: [configuration]
+Scope: [all/syntax/semantics]
+
+[Invoke multilingual]
+Language: [target language]
+```
diff --git a/hyperextract-skills/brainstorm/SKILL.md b/hyperextract-skills/brainstorm/SKILL.md
new file mode 100644
index 0000000..ec4e8d6
--- /dev/null
+++ b/hyperextract-skills/brainstorm/SKILL.md
@@ -0,0 +1,93 @@
+---
+name: hyper-extract-brainstorm
+description: |
+ Hyper-Extract Requirements Brainstorming Assistant. Explore user requirements and discuss detailed design for each type.
+
+ ## What We Explore
+ - Input: Source, domain, structure
+ - Output: What to extract, how to use
+ - Type-specific discussions:
+ * Record types: Item structure, field design, deduplication
+ * Graph types: Entity types, relation types, multi-type nodes
+ * Hypergraph: Participant types, outcome structure, reasoning needs
+ * Temporal/spatial: Time/location on edges, format handling, observation reference
+
+ ## Output
+ 1. Requirements summary
+ 2. Recommended type
+ 3. Type-specific design specs
+---
+
+# Brainstorm: Requirements Exploration
+
+## Phase 1: Input & Output
+
+**Input**: Source, domain, structure level, length
+**Output**: What to extract and how to use
+
+## Phase 2: Type Determination
+
+- "Complete summary" → **model**
+- "List of items" → **list**
+- "Deduplicated entities" → **set**
+- "Relationships between entities" → **graph**
+- "Multi-party relationships" → **hypergraph**
+- "Events with time" → **temporal_graph**
+- "Events with location" → **spatial_graph**
+- "Events with time + location" → **spatio_temporal_graph**
+
+## Phase 3: Type-Specific Discussion
+
+### Record Types (model/list/set)
+
+- Item structure and fields
+- For set: deduplication field
+- For list: nested structure, order
+
+### Graph Types
+
+1. **Entity types**: What entities to extract?
+2. **Entity fields**: name, category, description?
+3. **Relation types**: What relationships?
+4. **Relation fields**: Binary or complex?
+
+### Hypergraph
+
+- Participants: Who/what participates?
+- Outcome: What result/conclusion?
+- Reasoning: Evidence needed?
+
+### Dimension Evolution (temporal/spatial)
+
+After basic graph design, discuss:
+
+**Time dimension?**
+- Does the relation have time? (e.g., "A bought from B on date X")
+- Is time on node (birth_date) or edge (event_date)?
+- If time needs to be a relation dimension → upgrade to **temporal_graph**
+- Extract time to edge, add `time_field` in identifiers
+
+**Location dimension?**
+- Does the relation have location? (e.g., "A bought from B at location Y")
+- Is location on node (hq) or edge (transaction location)?
+- If location needs to be a relation dimension → upgrade to **spatial_graph**
+- Extract location to edge, add `location_field` in identifiers
+
+**Both?**
+- Need both time and location → **spatio_temporal_graph**
+
+## Output Format
+
+```markdown
+## Requirements Summary
+- Input: ...
+- Output: ...
+- Quality: ...
+
+## Type Recommendation
+**Type**: [type]
+**Confidence**: High/Medium/Low
+
+## Design Specs
+[Based on type discussion]
+```
diff --git a/hyperextract-skills/examples/graph/knowledge_graph.yaml b/hyperextract-skills/examples/graph/knowledge_graph.yaml
new file mode 100644
index 0000000..150ff0b
--- /dev/null
+++ b/hyperextract-skills/examples/graph/knowledge_graph.yaml
@@ -0,0 +1,68 @@
+language: zh
+name: knowledge_graph
+type: graph
+tags: [general]
+description: '通用知识图谱模板 - 从非结构化文本中提取关键实体、实体类型及实体之间的显式关系,适用于新闻、百科、制度和说明类文本的基础结构化建模。'
+
+output:
+ description: '由实体节点和关系边组成的通用知识图谱结果,用于表达文本中的核心对象及其语义连接。'
+ entities:
+ description: '文本中可独立识别、可作为关系端点的实体节点。'
+ fields:
+ - name: name
+ type: str
+ description: '实体在文本中的标准称呼或最稳定名称。'
+ - name: type
+ type: str
+ description: '实体类别,如人物、地点、组织、物品、概念、事件、制度、技术等。'
+ - name: description
+ type: str
+ description: '对实体身份、含义或在当前文本中作用的简要说明。'
+ required: false
+ relations:
+ description: '实体之间的显式语义关系边。'
+ fields:
+ - name: source
+ type: str
+ description: '关系起点实体,必须对应已提取实体的 name。'
+ - name: target
+ type: str
+ description: '关系终点实体,必须对应已提取实体的 name。'
+ - name: type
+ type: str
+ description: '关系类型,应使用简洁明确的语义词,如包含、属于、合作、影响、导致、相关等。'
+ - name: description
+ type: str
+ description: '对关系依据、方向或上下文含义的简要说明。'
+ required: false
+
+guideline:
+ target:
+ zh: '你是一位知识抽取与知识图谱构建专家。请基于给定文本识别关键实体,并构建仅由这些实体组成的二元关系图。'
+ rules_for_entities:
+ zh:
+ - '提取对理解文本事实、结构或逻辑有价值的实体,避免无意义的泛称和临时短语。'
+ - 'name 使用原文中最明确、最稳定的称呼;同一实体在全文中保持命名一致。'
+ - 'type 需概括实体类别,可使用人物、地点、组织、物品、概念、事件、制度、技术等标签。'
+ - 'description 用 1 句话概括实体的身份、属性或在文本中的作用;若信息不足可留空。'
+ rules_for_relations:
+ zh:
+ - '仅在文本明确表达两个实体之间存在语义联系时创建关系,不要凭常识补充隐含关系。'
+ - 'source 和 target 必须直接引用已提取实体的 name,不能新造实体名称。'
+ - 'type 使用简洁、方向清晰的关系词,如包含、属于、相关、影响、合作、竞争、导致、位于、使用等。'
+ - 'description 简要说明关系依据、作用方式或上下文;若文本没有额外信息可留空。'
+ - '避免重复关系;若同一对实体存在多种不同关系,可分别记录。'
+
+identifiers:
+ entity_id: name
+ relation_id: '{source}|{target}'
+ relation_members:
+ source: source
+ target: target
+
+options:
+ extraction_mode: two_stage
+
+display:
+ entity_label: '{name} ({type})'
+ relation_label: '{type}'
diff --git a/hyperextract-skills/examples/hypergraph/compliance_logic.yaml b/hyperextract-skills/examples/hypergraph/compliance_logic.yaml
new file mode 100644
index 0000000..047c6f2
--- /dev/null
+++ b/hyperextract-skills/examples/hypergraph/compliance_logic.yaml
@@ -0,0 +1,117 @@
+language: zh
+name: compliance_logic
+type: hypergraph
+tags: [general]
+description:
+ zh: '规定条文超图 - 面向制度、细则、办法、规范、指引、规程等规定类文本,提取条文中涉及的名词性要素,并用 action 概括这些要素之间被要求执行的行为。'
+
+output:
+ description:
+ zh: '由名词性实体节点和规定超边组成的结构化结果。节点只保留规定中涉及的人、部门、文件、场所、物品等要素;超边表示一条独立规定,action 用于概括这些要素之间被要求执行的行为。'
+ entities:
+ description:
+ zh: '规定涉及的基础节点。只提取名词或名词短语,不提取动作本身。'
+ fields:
+ - name: name
+ type: str
+ description:
+ zh: '实体名称,尽量直接使用原文中的稳定称呼;同一对象若同时出现全称、简称或缩写,应统一为一个名称。'
+ - name: category
+ type: str
+ description:
+ zh: '实体类型,应尽量使用清晰、稳定的名词类别,如人员、部门、组织、岗位、文件表单、证件、场所、设备、物品、时间要求、其他。'
+ - name: description
+ type: str
+ description:
+ zh: '用一句话说明该实体在当前规定中的含义或作用。'
+ required: false
+ default: ''
+ relations:
+ description:
+ zh: '表达规定条文的超边。每条超边对应一条独立规定,由相关实体构成,action 负责概括这些实体之间被要求执行的行为。'
+ fields:
+ - name: ruleType
+ type: str
+ description:
+ zh: '规定类型,用于概括该条规定在规定什么。优先使用:职责分工、行为要求、流程程序、其他。'
+ - name: participants
+ type: list[str]
+ description:
+ zh: '构成该条规定基本内容的实体名称列表,必须直接使用已提取节点的 name。只放实体,不要把动作短语放进 participants。'
+ - name: action
+ type: str
+ description:
+ zh: '该条规定涉及的核心行为概括,用 1 句话说明这些实体之间被要求执行什么行为。action 承担行为表达,不要把 action 再拆成实体。'
+ - name: sourceClause
+ type: str
+ description:
+ zh: '规定来源的条款编号、标题或出处标识,尽量写最近的编号或小节定位。'
+ required: false
+ default: ''
+
+guideline:
+ target:
+ zh: '你是一位规定条文结构化分析专家。请从规定类文本中提取每条规定涉及的名词性要素,并用 action 写清这些要素之间被要求执行的行为。'
+ rules_for_entities:
+ zh:
+ - '提取名词或名词短语形式的实体'
+ - '优先提取规定中真正参与行为的要素,如人员、岗位、部门、组织、文件表单、证件、场所、设备、物品、时间要求等。根据文本的具体内容和你的知识来决定具体的类别。'
+ - 'name 尽量保持原文中的稳定称呼;同一对象不要重复命名。'
+ - '如果某个词只是行为本身,例如“登记”“审批”“填写交接记录”,不要单独作为实体。'
+ - '对于“访客证”“登记表”“样品库”“前台”“每日18:00前”这类名词性短语,可以作为实体提取。'
+ - 'category 要尽量使用少数、容易理解的类别,不要为了细分而临时创造模糊类别。'
+ rules_for_relations:
+ zh:
+ - '每条超边只表达一条独立规定,并连接这条规定涉及的相关实体。'
+ - 'ruleType 优先使用:职责分工、行为要求、流程程序、其他。根据文本的具体内容和你的知识来决定具体的类别。'
+ - 'participants 必须直接引用已提取节点的 name,只放名词性实体,不放动作短语。'
+ - 'action 用 1 句话概括这条规定要求这些实体之间发生什么行为。'
+ - |
+ 完整示例(输入):
+ 1. 访客进入样品库前,应在前台登记并领取访客证。
+ 2. 安全部门每日18:00前回收当日访客证,并存档登记表。
+ - |
+ 完整示例(输出):
+ entities:
+ - name: 访客
+ category: 人员
+ description: 进入样品库并需履行登记要求的外来人员。
+ - name: 样品库
+ category: 场所
+ description: 访客进入前需要满足规定要求的区域。
+ - name: 前台
+ category: 场所
+ description: 办理登记和领取访客证的地点。
+ - name: 访客证
+ category: 证件
+ description: 访客登记后领取并在离开前回收的证件。
+ - name: 安全部门
+ category: 部门
+ description: 负责回收访客证并保存记录的管理部门。
+ - name: 登记表
+ category: 文件表单
+ description: 记录访客登记信息并需要存档的表单。
+ - name: 每日18:00前
+ category: 时间要求
+ description: 安全部门执行回收和存档要求的截止时间要求。
+ relations:
+ - ruleType: 流程程序
+ participants: [访客, 样品库, 前台, 访客证]
+ action: 规定访客进入样品库前,应在前台登记并领取访客证。
+ sourceClause: 1
+ - ruleType: 流程程序
+ participants: [安全部门, 访客证, 登记表, 每日18:00前]
+ action: 规定安全部门每日18:00前回收当日访客证,并存档登记表。
+ sourceClause: 2
+
+identifiers:
+ entity_id: name
+ relation_id: '{sourceClause}|{ruleType}|{action}'
+ relation_members: participants
+
+options:
+ extraction_mode: one_stage
+
+display:
+ entity_label: '{name} ({category})'
+ relation_label: '[{ruleType}] {action}'
diff --git a/hyperextract-skills/examples/list/clause_list.yaml b/hyperextract-skills/examples/list/clause_list.yaml
new file mode 100644
index 0000000..d78f9bb
--- /dev/null
+++ b/hyperextract-skills/examples/list/clause_list.yaml
@@ -0,0 +1,50 @@
+language: zh
+name: clause_list
+type: list
+tags: [general]
+description:
+ zh: '规章条文清单 - 将规章制度、操作手册或合规文本拆解为逐条的原子化条文,便于索引、检索、比对与问答。'
+
+output:
+ description:
+ zh: '由逐条拆解后的规章条文组成的列表结果。'
+ fields:
+ - name: clauseId
+ type: str
+ description:
+ zh: '条文编号,如第1条、第2条、第2.1款、第3项等。'
+ - name: title
+ type: str
+ description:
+ zh: '条文标题、主题或该条的简短概括;原文未明确给出时可留空。'
+ required: false
+ default: ''
+ - name: content
+ type: str
+ description:
+ zh: '条文正文内容,尽量保持与原文一致,避免改写。'
+ - name: category
+ type: str
+ description:
+ zh: '条文分类,如总则、权利义务、程序、处罚、附则、其他。'
+ required: false
+ default: '其他'
+
+guideline:
+ target:
+ zh: '你是一位专业的规章制度条文拆解专家。请将输入文本按其自然结构拆解为可独立引用、可检索的逐条条文。'
+ rules:
+ zh:
+ - '按原文的自然层级和编号结构拆解条文,如第X条、第X款、第X项等。'
+ - '每一项输出都应对应一个独立、可单独引用的条文单元。'
+ - 'clauseId 必须尽量使用原文中的编号;若原文无明确编号,可使用稳定且简洁的顺序标识。'
+ - 'content 应保留该条的核心正文,尽量与原文一致,不要总结式改写。'
+ - 'title 仅在原文明确存在标题、主题或能够从该条主旨中直接提炼时填写;否则留空。'
+ - 'category 需尽量归入总则、权利义务、程序、处罚、附则、其他之一。'
+ - '不要遗漏任何实质性条文,也不要把多个独立条文合并成一条。'
+ - '目录、页眉页脚、签章、编号噪音、纯格式性重复内容通常不作为条文输出。'
+
+options: {}
+
+display:
+ label: '{clauseId}({title})'
diff --git a/hyperextract-skills/examples/model/news_summary.yaml b/hyperextract-skills/examples/model/news_summary.yaml
new file mode 100644
index 0000000..a29a7a1
--- /dev/null
+++ b/hyperextract-skills/examples/model/news_summary.yaml
@@ -0,0 +1,32 @@
+language: zh
+name: NewsSummary
+type: model
+tags: [news]
+description: '新闻摘要 - 从新闻报道中提取关键信息'
+
+output:
+ description: '新闻报道的结构化摘要'
+ fields:
+ - name: title
+ type: str
+ description: '新闻标题'
+ - name: publication_date
+ type: str
+ description: '发布日期'
+ required: false
+ - name: source
+ type: str
+ description: '新闻来源'
+ required: false
+ - name: summary
+ type: str
+ description: '新闻摘要'
+
+guideline:
+ target: '你是一位新闻编辑。请从新闻报道中提取关键信息。'
+ rules:
+ - 'title 保留原文标题'
+ - 'summary 用 2-3 句话概括核心内容'
+
+display:
+ label: '{title}'
diff --git a/hyperextract-skills/examples/set/entity_registry.yaml b/hyperextract-skills/examples/set/entity_registry.yaml
new file mode 100644
index 0000000..2abb4e4
--- /dev/null
+++ b/hyperextract-skills/examples/set/entity_registry.yaml
@@ -0,0 +1,45 @@
+language: zh
+name: entity_registry
+type: set
+tags: [general]
+description:
+ zh: '实体注册表模板 - 从任意文本中识别并去重关键实体,汇总为可复用的实体清单。'
+
+output:
+ description:
+ zh: '去重后的实体列表结果,用于形成基础实体台账。'
+ fields:
+ - name: name
+ type: str
+ description:
+ zh: '实体名称,优先使用文本中最稳定、最清晰的称呼。'
+ - name: category
+ type: str
+ description:
+ zh: '实体类型,如人物、机构、地点、产品、概念、事件、制度、其他。'
+ - name: description
+ type: str
+ description:
+ zh: '对实体身份、含义或在文本中作用的简要说明。'
+ required: false
+ default: ''
+
+guideline:
+ target:
+ zh: '你是一位实体识别专家。请从文本中识别具有独立意义的实体,并将重复出现的同一实体合并为一条记录。'
+ rules:
+ zh:
+ - '提取对理解文本内容有意义的实体,如人物、机构、地点、产品、概念、事件、制度等。'
+ - 'name 使用文本中最清晰、最稳定的称呼;同一实体若有简称或别称,应统一为一个名称。'
+ - 'category 用于概括实体类型,尽量准确且可复用。'
+ - 'description 用 1 句话说明实体是谁、是什么或在文本中的作用;信息不足时可留空。'
+ - '对同一实体进行去重,不要因表述差异重复输出多条。'
+ - '不要把纯代词、泛称、临时描述短语或无独立指代意义的词语当作实体。'
+
+identifiers:
+ item_id: name
+
+options: {}
+
+display:
+ label: name
diff --git a/hyperextract-skills/examples/spatial_graph/purchase_record.yaml b/hyperextract-skills/examples/spatial_graph/purchase_record.yaml
new file mode 100644
index 0000000..24d1908
--- /dev/null
+++ b/hyperextract-skills/examples/spatial_graph/purchase_record.yaml
@@ -0,0 +1,54 @@
+language: zh
+name: PurchaseRecord
+type: spatial_graph
+tags: [commerce]
+description: '购买记录图 - 记录购买关系及购买发生地点'
+
+output:
+ description: '购买关系图,关系边包含购买地点'
+ entities:
+ description: '买家、卖家和商品节点'
+ fields:
+ - name: name
+ type: str
+ description: '实体名称'
+ - name: type
+ type: str
+ description: '实体类型:买家、卖家、商品'
+ relations:
+ description: '购买关系边'
+ fields:
+ - name: source
+ type: str
+ description: '源实体(买家)'
+ - name: target
+ type: str
+ description: '目标实体(商品)'
+ - name: amount
+ type: str
+ description: '购买金额'
+ required: false
+ - name: location
+ type: str
+ description: '购买发生地点'
+
+guideline:
+ target: '你是一位商业分析专家。请提取购买关系和购买地点信息。'
+ rules_for_entities:
+ - '提取买家、卖家和商品'
+ rules_for_relations:
+ - 'source 必须是买家,target 必须是商品'
+ rules_for_location:
+ - 'location 记录购买发生的地点'
+
+identifiers:
+ entity_id: name
+ relation_id: '{source}|{target}'
+ relation_members:
+ source: source
+ target: target
+ location_field: location
+
+display:
+ entity_label: '{name} ({type})'
+ relation_label: '购买地点: {location}'
diff --git a/hyperextract-skills/examples/spatio_temporal_graph/social_meeting.yaml b/hyperextract-skills/examples/spatio_temporal_graph/social_meeting.yaml
new file mode 100644
index 0000000..476b88a
--- /dev/null
+++ b/hyperextract-skills/examples/spatio_temporal_graph/social_meeting.yaml
@@ -0,0 +1,61 @@
+language: zh
+name: SocialMeeting
+type: spatio_temporal_graph
+tags: [social]
+description: '社交会面图 - 记录人物之间的会面关系,包含会面时间和地点'
+
+output:
+ description: '社交会面关系图,关系边包含会面时间和地点'
+ entities:
+ description: '人物节点'
+ fields:
+ - name: name
+ type: str
+ description: '人物姓名'
+ - name: role
+ type: str
+ description: '社会角色'
+ required: false
+ relations:
+ description: '会面关系边'
+ fields:
+ - name: source
+ type: str
+ description: '源实体(人物A)'
+ - name: target
+ type: str
+ description: '目标实体(人物B)'
+ - name: meeting_type
+ type: str
+ description: '会面类型'
+ required: false
+ - name: event_date
+ type: str
+ description: '会面时间'
+ - name: location
+ type: str
+ description: '会面地点'
+
+guideline:
+ target: '你是一位社交分析专家。请提取人物之间的会面关系。'
+ rules_for_entities:
+ - '提取所有涉及的人物'
+ rules_for_relations:
+ - 'source 和 target 是会面的双方'
+ rules_for_time:
+ - 'event_date 记录会面时间'
+ rules_for_location:
+ - 'location 记录会面地点'
+
+identifiers:
+ entity_id: name
+ relation_id: '{source}|{target}|{event_date}'
+ relation_members:
+ source: source
+ target: target
+ time_field: event_date
+ location_field: location
+
+display:
+ entity_label: '{name}'
+ relation_label: '{event_date} @ {location}'
diff --git a/hyperextract-skills/examples/temporal_graph/biographical_timeline.yaml b/hyperextract-skills/examples/temporal_graph/biographical_timeline.yaml
new file mode 100644
index 0000000..ccc5902
--- /dev/null
+++ b/hyperextract-skills/examples/temporal_graph/biographical_timeline.yaml
@@ -0,0 +1,104 @@
+language: zh
+name: biographical_timeline
+type: temporal_graph
+tags: [general]
+description:
+ zh: '人物生平时间线 - 从传记、自传、回忆录等人物叙事文本中提取带时间的生平事件,并组织为时间线。'
+
+output:
+ description:
+ zh: '由人物相关实体和生平事件关系组成的时间线结果,用于还原人物经历、角色变化与关键转折。'
+ entities:
+ description:
+ zh: '文本中与人物生平相关、可作为事件端点的实体。'
+ fields:
+ - name: name
+ type: str
+ description:
+ zh: '实体在文本中的稳定名称或常用称呼。'
+ - name: type
+ type: str
+ description:
+ zh: '实体类型,如人物、地点、组织、作品、奖项、身份、物品、概念、其他。'
+ - name: description
+ type: str
+ description:
+ zh: '对实体身份、角色或在当前文本中作用的简要说明。'
+ required: false
+ relations:
+ description:
+ zh: '围绕人物经历构建的事件关系边。'
+ fields:
+ - name: source
+ type: str
+ description:
+ zh: '事件发起者或经历者,通常为人物主体或触发事件的相关实体。'
+ - name: target
+ type: str
+ description:
+ zh: '与该事件直接相关的对象实体,如地点、学校、组织、人物、作品、奖项等。'
+ - name: eventType
+ type: str
+ description:
+ zh: '事件类型。使用能够概括事件本身的简洁词语或短语,优先具体、直观、可区分;通常可用动词,如出生、入学、任职、结婚、出版、获奖。可参考示例,但不必局限于示例,也可以根据文本语义自行创造更贴切的标签。'
+ - name: eventDate
+ type: str
+ description:
+ zh: '事件时间。优先保留文本中最精确的时间表达;可为年份、年月、年月日,必要时可保留时间段。'
+ required: false
+ - name: details
+ type: str
+ description:
+ zh: '事件补充说明,可记录身份、地点、机构、结果或上下文。'
+ required: false
+
+guideline:
+ target:
+ zh: '你是一位专业的人物传记事件抽取专家。请围绕人物主体,从文本中提取与其生平经历直接相关的实体和事件,并构建可信的时间线。'
+ rules_for_entities:
+ zh:
+ - '优先提取与人物主体生平经历直接相关的实体,如人物、地点、组织、作品、奖项、身份称谓等。'
+ - '为每个实体指定尽量准确的类型,并优先使用文本中的稳定称呼。'
+ - '不必为纯背景概念、空泛评价或无法成为事件端点的内容单独建实体。'
+ rules_for_relations:
+ zh:
+ - '仅从已提取实体中创建事件边,source 和 target 都必须对应实体 name。'
+ - '事件应代表可放入人物时间线的经历、转折、公开活动或身份变化,不要把纯评价、性格描述或静态背景知识当作事件。'
+ - '若一句话包含多个独立事件,应拆分为多条关系;若同一事件同时涉及时间、地点、组织、身份,可在 details 中补充。'
+ - 'eventType 应使用能够概括事件本身的简洁词语或短语,优先具体、直观、可区分;通常可用动词或动宾结构。'
+ - '优先写贴近原文语义的具体标签,不要过度抽象。例如优先写“入学、毕业、留学”而不是“教育”,优先写“任职、升迁、离任”而不是“工作”,优先写“出版、获奖、发明”而不是“成就”。'
+ - '可参考但不限于以下示例:出生、逝世、迁居、入学、毕业、任职、结婚、合作、出版、获奖、参战、归国、隐居。'
+ - '允许根据文本语义自行创造更贴切的标签,例如拜师、出道、登基、封侯、退赛、复出等;不要求必须使用示例中的词。'
+ - '仅在确实无法提炼出更具体的事件标签时,才使用较泛的词。'
+ - '时间信息为可选,仅在文本明确给出或能根据明确相对时间可靠换算时才填写。'
+ rules_for_time:
+ zh:
+ - '当前观察日期:{observation_time}'
+ - '相对时间解析(基于观察日期):'
+ - '- "去年" → {observation_time} 前一年'
+ - '- "今年" → {observation_time} 所属年份'
+ - '- "明年" → {observation_time} 后一年'
+ - '- "昨天/今天/明天" → {observation_time} 前/当/后一天'
+ - '- "十年前/五年后" → {observation_time} 前/后相应年数'
+ - '- "年初/年末" → 当年 1月1日/12月31日'
+ - '- "朝代" → 保留朝代名称'
+ - '明确时间:优先保留文本中最精确的原始格式'
+ - '- 年份(如 429年、2023年)→ 429、2023'
+ - '- 年-月、年-月-日、时间段(如 1937-1945)→ 保持原格式'
+ - '若只有模糊阶段性表达且无法可靠换算,不要强行补时间'
+ - '缺失时间:留空'
+
+identifiers:
+ entity_id: name
+ relation_id: '{source}|{target}'
+ relation_members:
+ source: source
+ target: target
+ time_field: 'eventDate'
+
+options:
+ extraction_mode: two_stage
+
+display:
+ entity_label: '{name} ({type})'
+ relation_label: '{eventType}({eventDate})'
diff --git a/hyperextract-skills/graph-designer/SKILL.md b/hyperextract-skills/graph-designer/SKILL.md
new file mode 100644
index 0000000..5368c22
--- /dev/null
+++ b/hyperextract-skills/graph-designer/SKILL.md
@@ -0,0 +1,358 @@
+---
+name: hyper-extract-graph-designer
+description: |
+ Hyper-Extract Graph Designer. Designs structure, identifiers, display, and extraction rules for graph/hypergraph/temporal/spatial types.
+
+ ## Applicable Types
+ - graph: Binary relations
+ - hypergraph: Multi-entity relations
+ - temporal_graph: + time dimension
+ - spatial_graph: + space dimension
+ - spatio_temporal_graph: + time + space
+
+ ## Complete Output
+ - output.entities
+ - output.relations
+ - guideline
+ - identifiers
+ - display
+---
+
+# Graph Designer: graph/hypergraph/temporal/spatial
+
+## Type Confirmation
+
+- **graph**: Simple binary relations (A→B)
+- **hypergraph**: Multi-party relations (A+B+C→D)
+- **temporal_graph**: Binary relations + time
+- **spatial_graph**: Binary relations + space
+- **spatio_temporal_graph**: Binary relations + time + space
+
+## Part 1: Entity Design (output.entities)
+
+### Standard Entity Fields
+
+```yaml
+entities:
+ description: 'Node/entity definitions'
+ fields:
+ - name: name
+ type: str
+ description: 'Entity name (unique identifier)'
+ - name: category
+ type: str
+ description: 'Entity type/category'
+ - name: description
+ type: str
+ description: 'Brief description'
+ required: false
+ default: ''
+```
+
+### Multi-Type Node Considerations
+
+If nodes can have multiple categories:
+
+```yaml
+ - name: category
+ type: list
+ description: 'Entity types (can be multiple)'
+```
+
+### Domain-Specific Fields
+
+Add fields based on your domain:
+
+```yaml
+ - name: role
+ type: str
+ description: 'Role in current context'
+ required: false
+ default: ''
+```
+
+## Part 2: Relation Design (output.relations)
+
+### For graph/temporal/spatial (Binary Relations)
+
+```yaml
+relations:
+ description: 'Edge/relation definitions'
+ fields:
+ - name: source
+ type: str
+ description: 'Source entity name'
+ - name: target
+ type: str
+ description: 'Target entity name'
+ - name: relation_type
+ type: str
+ description: 'Type of relationship'
+```
+
+### Predefined Relation Types
+
+Common types (can be customized):
+
+| Category | Relation Types |
+|----------|---------------|
+| Social | owns, works_for, married_to, parent_of |
+| Organizational | subsidiary_of, competitor_of, partner_with |
+| Causal | causes, enables, prevents, leads_to |
+| Spatial | located_at, adjacent_to, part_of |
+| Temporal | preceded_by, followed_by, concurrent_with |
+
+### For hypergraph (Multi-Entity Relations)
+
+```yaml
+relations:
+ description: 'Hyperedge definitions'
+ fields:
+ - name: relation_type
+ type: str
+ description: 'Type of relationship'
+ - name: participants
+ type: list
+ description: 'List of participating entity names'
+ - name: outcome
+ type: str
+ description: 'Result or conclusion'
+ - name: reasoning
+ type: str
+ description: 'Explanation or evidence'
+ required: false
+ default: ''
+```
+
+### For temporal_graph (Add time field on edges)
+
+```yaml
+ - name: event_date
+ type: str
+ description: 'When the relation occurred'
+ required: false
+ default: ''
+```
+
+### For spatial_graph (Add location field on edges)
+
+```yaml
+ - name: location
+ type: str
+ description: 'Where the relation occurred'
+ required: false
+ default: ''
+```
+
+## Part 3: Extraction Rules (guideline)
+
+### Target Setting
+
+```yaml
+guideline:
+ target: 'You are a [domain] expert. Extract knowledge graph from text.'
+```
+
+### Entity Extraction Rules
+
+```yaml
+ rules_for_entities:
+ - 'Entity names must match exactly with text'
+ - 'Each entity must have a clear category'
+ - 'Only extract entities explicitly mentioned'
+```
+
+### Relation Extraction Rules
+
+```yaml
+ rules_for_relations:
+ - 'Only create edges for explicitly stated relations'
+ - 'source and target must reference extracted entities'
+ - 'Use predefined relation types when possible'
+```
+
+### For temporal_graph (Time handling rules)
+
+```yaml
+ rules_for_time:
+ - 'Observation time: {observation_time}'
+ - 'Absolute dates: Keep as-is (e.g., 2024-01-01)'
+ - 'Relative time: Convert to absolute (last year → previous year from observation)'
+ - 'Fuzzy time: Leave empty, do not guess'
+```
+
+### For spatial_graph (Location handling rules)
+
+```yaml
+ rules_for_location:
+ - 'Observation location: {observation_location}'
+ - 'Structured locations: Keep as-is (e.g., Beijing, China)'
+ - 'Fuzzy locations: Use observation_location (here → observation)'
+ - 'Coordinates: Keep as-is'
+```
+
+## Part 4: Identifier Rules (identifiers)
+
+### For binary relations (graph/temporal/spatial)
+
+```yaml
+identifiers:
+ entity_id: name
+ relation_id: '{source}|{relation_type}|{target}'
+ relation_members:
+ source: source
+ target: target
+```
+
+### For hypergraph
+
+The `relation_members` is a **string** pointing to the participants field name (type: list).
+
+```yaml
+identifiers:
+ entity_id: name
+ relation_id: '{ruleType}|{action}'
+ relation_members: participants # STRING, points to list field
+```
+
+### For complex cases (relation_members as list)
+
+When a relation involves multiple sources of participants:
+
+```yaml
+identifiers:
+ entity_id: name
+ relation_id: '{relation_type}'
+ relation_members: [source_participants, target_participants]
+```
+
+The system will unpack the lists and use the combined set as members.
+
+### For temporal_graph (Add time identifier)
+
+```yaml
+ time_field: event_date
+```
+
+### For spatial_graph (Add location identifier)
+
+```yaml
+ location_field: location
+```
+
+### Identifier Template Syntax
+
+- `{field_name}`: Insert field value
+- `{source}`: Source entity name
+- `{relation_type}`: Relation type
+- `{target}`: Target entity name
+
+## Part 5: Display Configuration (display)
+
+### For binary relations
+
+```yaml
+display:
+ entity_label: '{name} ({category})'
+ relation_label: '{relation_type}'
+```
+
+### For hypergraph
+
+```yaml
+display:
+ entity_label: '{name} ({category})'
+ relation_label: '[{ruleType}] {outcome}'
+```
+
+## Design Checklist
+
+### Entities
+- [ ] Entity types cover key concepts?
+- [ ] Entity granularity appropriate?
+- [ ] Multi-type nodes handled correctly?
+- [ ] No redundant fields?
+
+### Relations
+- [ ] Relation types semantically clear?
+- [ ] source/target reference defined entities?
+- [ ] No ambiguous relations?
+
+### Hypergraph Specific
+- [ ] Participant count reasonable?
+- [ ] Outcome clearly defined?
+- [ ] Reasoning/evidence needed?
+
+### Identifiers
+- [ ] entity_id references name field?
+- [ ] relation_id template correct?
+- [ ] relation_members configured correctly (string for hypergraph)?
+- [ ] time_field/location_field specified for temporal/spatial?
+
+## Complete Output Example
+
+```yaml
+language: en
+
+name: CompanyRelationshipGraph
+type: graph
+tags: [finance, corporate]
+description: 'Extract company ownership and relationship graphs'
+
+output:
+ entities:
+ description: 'Company and person entities'
+ fields:
+ - name: name
+ type: str
+ description: 'Entity name'
+ - name: category
+ type: str
+ description: 'Entity type: Company, Person'
+ - name: description
+ type: str
+ description: 'Brief description'
+ required: false
+ default: ''
+ relations:
+ description: 'Ownership and control relationships'
+ fields:
+ - name: source
+ type: str
+ description: 'Source entity'
+ - name: target
+ type: str
+ description: 'Target entity'
+ - name: relation_type
+ type: str
+ description: 'owns, controls, invested_in'
+
+guideline:
+ target: 'You are a corporate analysis expert. Extract company relationships.'
+ rules_for_entities:
+ - 'Extract company and person entities only'
+ - 'Category must be Company or Person'
+ - 'Names must match text exactly'
+ rules_for_relations:
+ - 'Only extract ownership and control relationships'
+ - 'Common types: owns, controls, invested_in, subsidiary_of'
+ - 'source and target must be extracted entities'
+
+identifiers:
+ entity_id: name
+ relation_id: '{source}|{relation_type}|{target}'
+ relation_members:
+ source: source
+ target: target
+
+display:
+ entity_label: '{name} ({category})'
+ relation_label: '{relation_type}'
+```
+
+## Next Steps
+
+After graph design:
+1. Review with user
+2. Optional: yaml-validator
+3. Optional: multilingual
diff --git a/hyperextract-skills/multilingual/SKILL.md b/hyperextract-skills/multilingual/SKILL.md
new file mode 100644
index 0000000..d43d028
--- /dev/null
+++ b/hyperextract-skills/multilingual/SKILL.md
@@ -0,0 +1,119 @@
+---
+name: hyper-extract-multilingual
+description: |
+ Hyper-Extract Multilingual Converter. Converts single-language configuration to multi-language support.
+
+ ## Translatable Fields
+ - description
+ - output.description
+ - fields[].description
+ - guideline.target
+ - guideline.rules
+
+ ## Non-Translatable Fields
+ - name, type, tags
+ - field names (name in fields)
+ - identifiers
+ - display
+---
+
+# Multilingual Converter
+
+## Conversion Modes
+
+### Mode 1: Single → Bilingual
+
+```yaml
+# Before (single language)
+language: zh
+description: '财务报告摘要'
+
+# After (bilingual)
+language: [zh, en] # YAML list format
+description:
+ zh: '财务报告摘要'
+ en: 'Financial Report Summary'
+```
+
+### Mode 2: Expand Existing Translation
+
+```yaml
+# Before (has zh only)
+description:
+ zh: '财务报告摘要'
+
+# After (add en)
+description:
+ zh: '财务报告摘要'
+ en: 'Financial Report Summary'
+```
+
+### Mode 3: Add New Language
+
+```yaml
+# Before (zh + en)
+# After (add ja)
+description:
+ zh: '财务报告摘要'
+ en: 'Financial Report Summary'
+ ja: '財務報告サマリー'
+```
+
+## Translation Guidelines
+
+### Field Descriptions
+
+| Original | English | Notes |
+|----------|---------|-------|
+| 公司名称 | Company name | Direct |
+| 营收金额 | Revenue amount | Direct |
+| 关键词 | Keyword/Term | Choose based on context |
+
+### Target/Rules
+
+```yaml
+target:
+ zh: '你是一位财务分析师,请从财报中提取关键信息。'
+ en: 'You are a financial analyst. Extract key information from financial reports.'
+
+rules:
+ zh:
+ - '每个字段对应一个独立的信息项'
+ en:
+ - 'Each field corresponds to an independent information item'
+```
+
+## Translation Checklist
+
+- [ ] Preserve technical terms consistently
+- [ ] Maintain same meaning across languages
+- [ ] Keep field names in source language
+- [ ] Update language field to indicate multi-language (use list format)
+
+## Output Format
+
+```yaml
+## Multilingual Conversion Result
+
+language: [zh, en] # YAML list format
+
+# All translatable fields now have both zh and en
+
+## Translation Mapping
+
+| Field Path | Chinese | English |
+|------------|---------|---------|
+| description | ... | ... |
+| output.description | ... | ... |
+| fields[].description | ... | ... |
+| guideline.target | ... | ... |
+| guideline.rules | ... | ... |
+```
+
+## Notes
+
+- Use professional domain terminology
+- Maintain consistent terminology across fields
+- Keep YAML structure clean and readable
+- Field names (name) should NOT be translated
+- Use YAML list format for language field: `language: [zh, en]`
diff --git a/hyperextract-skills/record-designer/SKILL.md b/hyperextract-skills/record-designer/SKILL.md
new file mode 100644
index 0000000..24266d4
--- /dev/null
+++ b/hyperextract-skills/record-designer/SKILL.md
@@ -0,0 +1,168 @@
+---
+name: hyper-extract-record-designer
+description: |
+ Hyper-Extract Record Designer. Designs structure, identifiers, display, and extraction rules for model/list/set types.
+
+ ## Applicable Types
+ - model: Single object
+ - list: Object list
+ - set: Deduplicated set
+
+ ## Complete Output
+ - output.fields
+ - guideline (target + rules)
+ - identifiers.item_id (set only)
+ - display.label
+---
+
+# Record Designer: model/list/set
+
+## Type Confirmation
+
+- **model**: Extract a single complete object
+- **list**: Extract list of homogeneous objects
+- **set**: Extract deduplicated set of objects
+
+## Part 1: Field Design (output.fields)
+
+### Field Definition Template
+
+```yaml
+output:
+ description: 'Output structure description'
+ fields:
+ - name: field_name # snake_case, must be unique
+ type: str/int/float/list # data type
+ description: 'What this field means'
+ required: true/false # is this mandatory?
+ default: 'value' # default when missing
+```
+
+### Field Type Guidelines
+
+| Type | Use When | Example |
+|------|----------|---------|
+| str | Text values | names, descriptions |
+| int | Whole numbers | counts, years |
+| float | Decimal numbers | prices, percentages |
+| list | Multiple values | tags, categories |
+
+### Usually Needed Fields
+
+- At minimum, most records need a `name` or `title` field
+
+### Optional Fields
+
+- description: Brief explanation
+- category: Type or classification
+- Domain-specific fields
+
+## Part 2: Extraction Rules (guideline)
+
+### Target Setting
+
+```yaml
+guideline:
+ target: 'You are a [domain] expert. Extract [goal] from text.'
+```
+
+### Rules for Records
+
+```yaml
+ rules:
+ - 'Rule 1: What to extract'
+ - 'Rule 2: How to handle edge cases'
+ - 'Rule 3: Quality standards'
+```
+
+### Rule Writing Guidelines
+
+1. Be specific about what to extract
+2. Include edge case handling
+3. Set quality standards
+4. Prioritize important rules first
+
+## Part 3: Identifier Rules (identifiers)
+
+### For set types, configure deduplication:
+
+```yaml
+identifiers:
+ item_id: field_name # Deduplicate based on this field
+```
+
+### For model/list types:
+
+Usually not needed, or use empty:
+
+```yaml
+identifiers: {}
+```
+
+## Part 4: Display Configuration (display)
+
+```yaml
+display:
+ label: '{field_name}' # How to display each record
+```
+
+### Label Template Syntax
+
+- `{field_name}`: Insert field value
+- Example: `'{company_name}'` → "Acme Corp"
+
+## Design Checklist
+
+- [ ] All fields have clear semantic meaning?
+- [ ] Field types are appropriate (str/int/float/list)?
+- [ ] Required vs optional is reasonable?
+- [ ] Default values are safe/meaningful?
+- [ ] For set: item_id can uniquely identify records?
+- [ ] Display label references correct fields?
+
+## Complete Output Example
+
+```yaml
+language: en
+
+name: EarningsCallSummary
+type: model
+tags: [finance]
+description: 'Extract key financial metrics from earnings call transcripts'
+
+output:
+ description: 'Earnings call summary'
+ fields:
+ - name: company_name
+ type: str
+ description: 'Company name'
+ required: true
+ default: ''
+ - name: revenue
+ type: str
+ description: 'Revenue amount'
+ required: false
+ default: 'N/A'
+ - name: quarter
+ type: str
+ description: 'Fiscal quarter'
+ required: true
+ default: ''
+
+guideline:
+ target: 'You are a financial analyst. Extract key metrics from earnings calls.'
+ rules:
+ - 'Extract revenue in the format shown in text (including currency)'
+ - 'Use N/A if revenue is not mentioned'
+ - 'Extract quarter as Q1/Q2/Q3/Q4 format'
+
+display:
+ label: '{company_name} - {quarter}'
+```
+
+## Next Steps
+
+After record design:
+1. Review with user
+2. Optional: yaml-validator
+3. Optional: multilingual
diff --git a/hyperextract-skills/yaml-validator/SKILL.md b/hyperextract-skills/yaml-validator/SKILL.md
new file mode 100644
index 0000000..871b656
--- /dev/null
+++ b/hyperextract-skills/yaml-validator/SKILL.md
@@ -0,0 +1,173 @@
+---
+name: hyper-extract-yaml-validator
+description: |
+ Hyper-Extract YAML Validator. Checks configuration completeness, correctness, and best practices.
+
+ ## Validation Levels
+ - ERROR: Must fix, otherwise won't work
+ - WARNING: Should fix, may affect quality
+ - INFO: Reference info, recommended to follow
+---
+
+# YAML Validator
+
+## Validation Checklist
+
+### 1. Required Fields
+
+- [ ] language
+- [ ] name
+- [ ] type (one of 8 valid types)
+- [ ] tags
+- [ ] description
+- [ ] output
+- [ ] guideline
+
+### 2. Type-Specific Validation
+
+#### Record Types
+
+```yaml
+output:
+ description: '...' # ✓ Exists
+ fields: # ✓ Exists
+ - name: ... # ✓ Each field has name
+ type: ... # ✓ Has type
+ description: ... # ✓ Has description
+```
+
+#### Graph Types
+
+```yaml
+output:
+ entities: # ✓ Exists
+ description: '...'
+ fields:
+ - name: name
+ - name: category
+ relations: # ✓ Exists
+ description: '...'
+ fields:
+ - name: source # ✓ Binary relations
+ - name: target # ✓ OR
+ - name: participants # ✓ Hypergraph
+```
+
+#### Temporal Graph
+
+```yaml
+identifiers:
+ time_field: '...' # ✓ Must exist
+```
+
+#### Spatial Graph
+
+```yaml
+identifiers:
+ location_field: '...' # ✓ Must exist
+```
+
+### 3. Field Value Validation
+
+| Field | Valid Values | Common Errors |
+|-------|-------------|---------------|
+| type | model, list, set, graph, hypergraph, temporal_graph, spatial_graph, spatio_temporal_graph | Typo: "graphh" |
+| language | Single: zh/en; Multi: [zh, en] | Non-standard: "chinese" |
+| extraction_mode | one_stage, two_stage | Typo: "two-stage" |
+
+### 4. Naming Conventions
+
+- [ ] name: PascalCase (e.g., EarningsCallSummary)
+- [ ] tags: lowercase (e.g., [finance, medicine])
+- [ ] field names: snake_case (e.g., company_name)
+- [ ] relation_type: Usually Chinese (e.g., 拥有, 影响)
+
+### 5. Identifiers Configuration
+
+#### Binary Relations
+
+```yaml
+identifiers:
+ entity_id: name # ✓ References entity field
+ relation_id: '{source}|...' # ✓ Template syntax correct
+ relation_members: # ✓ Map correctly
+ source: source
+ target: target
+```
+
+#### Hypergraph
+
+```yaml
+identifiers:
+ entity_id: name
+ relation_id: '{ruleType}|{action}'
+ relation_members: participants # ✓ STRING, not dict
+```
+
+#### Complex (relation_members as list)
+
+```yaml
+identifiers:
+ entity_id: name
+ relation_id: '{relation_type}'
+ relation_members: [field1, field2] # ✓ List of field names
+```
+
+## Error Messages
+
+### Missing Required Field
+
+```
+ERROR: Missing required field 'output'
+Fix: Add output field definition
+```
+
+### Invalid Type Value
+
+```
+ERROR: type value 'graphh' is not valid
+Valid values: model, list, set, graph, hypergraph, temporal_graph, spatial_graph, spatio_temporal_graph
+Fix: Correct to 'graph'
+```
+
+### Missing Identifiers
+
+```
+ERROR: graph type requires identifiers.relation_members
+Fix: Add relation_members configuration
+```
+
+## Output Format
+
+```markdown
+## Validation Results
+
+### Syntax Validation
+✅ PASSED
+
+### Structure Validation
+✅ PASSED
+- language: ✅
+- name: ✅
+- type: ✅
+- tags: ✅
+- description: ✅
+- output: ✅
+
+### Semantic Validation
+⚠️ 2 warnings
+- WARNING: 'company_name' field description too brief
+- WARNING: Missing options.extraction_mode, recommend 'two_stage'
+
+### Best Practices
+⚠️ 1 suggestion
+- INFO: Consider adding identifiers for better deduplication
+
+### Overall Assessment
+✅ Configuration valid, ready for Hyper-Extract
+
+## Fix Suggestions
+
+1. Add extraction_mode: 'two_stage' to options
+2. Expand 'company_name' description to be more specific
+```