SKILL.md
只读
名称
ontology
描述
类型化知识图谱,用于结构化智能体记忆和可组合技能。在创建/查询实体(Person、Project、Task、Event、Document)、链接相关对象、强制执行约束、将多步操作规划为图转换,或技能需要共享状态时使用。触发词包括“记住”、“我知道关于X的什么”、“链接X到Y”、“显示依赖关系”、实体CRUD或跨技能数据访问。
本体(Ontology)
一个类型化词汇表+约束系统,用于将知识表示为可验证的图。
核心概念
一切都是具有类型、属性和与其他实体关系的实体。每次变更在提交前都会根据类型约束进行验证。
Entity: { id, type, properties, relations, created, updated }
Relation: { from_id, relation_type, to_id, properties }
何时使用
| 触发词 | 操作 |
|---|---|
| “记住...” | 创建/更新实体 |
| “我知道关于X的什么?” | 查询图 |
| “链接X到Y” | 创建关系 |
| “显示项目Z的所有任务” | 图遍历 |
| “什么依赖于X?” | 依赖查询 |
| 规划多步工作 | 建模为图转换 |
| 技能需要共享状态 | 读/写本体对象 |
核心类型
# 智能体与人员
Person: { name, email?, phone?, notes? }
Organization: { name, type?, members[] }
# 工作
Project: { name, status, goals[], owner? }
Task: { title, status, due?, priority?, assignee?, blockers[] }
Goal: { description, target_date?, metrics[] }
# 时间与地点
Event: { title, start, end?, location?, attendees[], recurrence? }
Location: { name, address?, coordinates? }
# 信息
Document: { title, path?, url?, summary? }
Message: { content, sender, recipients[], thread? }
Thread: { subject, participants[], messages[] }
Note: { content, tags[], refs[] }
# 资源
Account: { service, username, credential_ref? }
Device: { name, type, identifiers[] }
Credential: { service, secret_ref } # 切勿直接存储密钥
# 元数据
Action: { type, target, timestamp, outcome? }
Policy: { scope, rule, enforcement }
存储
默认路径:memory/ontology/graph.jsonl
{"op":"create","entity":{"id":"p_001","type":"Person","properties":{"name":"Alice"}}}
{"op":"create","entity":{"id":"proj_001","type":"Project","properties":{"name":"Website Redesign","status":"active"}}}
{"op":"relate","from":"proj_001","rel":"has_owner","to":"p_001"}
通过脚本或直接文件操作进行查询。对于复杂图,可迁移至SQLite。
工作流
创建实体
python3 scripts/ontology.py create --type Person --props '{"name":"Alice","email":"alice@example.com"}'
查询
python3 scripts/ontology.py query --type Task --where '{"status":"open"}'
python3 scripts/ontology.py get --id task_001
python3 scripts/ontology.py related --id proj_001 --rel has_task
链接实体
python3 scripts/ontology.py relate --from proj_001 --rel has_task --to task_001
验证
python3 scripts/ontology.py validate # 检查所有约束
约束
定义在 memory/ontology/schema.yaml 中:
types:
Task:
required: [title, status]
status_enum: [open, in_progress, blocked, done]
Event:
required: [title, start]
validate: "end >= start if end exists"
Credential:
required: [service, secret_ref]
forbidden_properties: [password, secret, token] # 强制间接引用
relations:
has_owner:
from_types: [Project, Task]
to_types: [Person]
cardinality: many_to_one
blocks:
from_types: [Task]
to_types: [Task]
acyclic: true # 无循环依赖
技能契约
使用本体的技能应声明:
# 在 SKILL.md 的 frontmatter 或头部
ontology:
reads: [Task, Project, Person]
writes: [Task, Action]
preconditions:
- "Task.assignee 必须存在"
postconditions:
- "创建的 Task 状态为 open"
作为图转换的规划
将多步计划建模为一系列图操作:
计划:“安排团队会议并创建后续任务”
1. CREATE Event { title: "团队同步", attendees: [p_001, p_002] }
2. RELATE Event -> has_project -> proj_001
3. CREATE Task { title: "准备议程", assignee: p_001 }
4. RELATE Task -> for_event -> event_001
5. CREATE Task { title: "发送总结", assignee: p_001, blockers: [task_001] }
每一步在执行前都经过验证。违反约束时回滚。
集成模式
与因果推断结合
将本体变更记录为因果动作:
# 创建/更新实体时,同时记录到因果动作日志
action = {
"action": "create_entity",
"domain": "ontology",
"context": {"type": "Task", "project": "proj_001"},
"outcome": "created"
}
跨技能通信
# 邮件技能创建承诺
commitment = ontology.create("Commitment", {
"source_message": msg_id,
"description": "周五前发送报告",
"due": "2026-01-31"
})
# 任务技能拾取
pending = ontology.query("Commitment", {"status": "pending"})
for c in pending:
ontology.create("Task", {
"title": c.description,
"due": c.due,
"source": c.id
})
快速开始
# 初始化本体存储
mkdir -p memory/ontology
touch memory/ontology/graph.jsonl
# 创建模式(可选但推荐)
cat > memory/ontology/schema.yaml << 'EOF'
types:
Task:
required: [title, status]
Project:
required: [name]
Person:
required: [name]
EOF
# 开始使用
python3 scripts/ontology.py create --type Person --props '{"name":"Alice"}'
python3 scripts/ontology.py list --type Person
参考
references/schema.md— 完整类型定义和约束模式references/queries.md— 查询语言和遍历示例






