KL8-2026 memory-ontology

Use when: need to store structured memory, remember people/projects/tasks/entities, recall context from previous sessions, build knowledge graph, manage ontology, track entities and their relations. Invokes the 'memory' MCP server to persist knowledge between conversations.

install
source · Clone the upstream repo
git clone https://github.com/meteor-007/KL8-2026
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/meteor-007/KL8-2026 "$T" && mkdir -p ~/.claude/skills && cp -r "$T/.github/skills/memory-ontology" ~/.claude/skills/meteor-007-kl8-2026-memory-ontology && rm -rf "$T"
manifest: .github/skills/memory-ontology/SKILL.md
source content

Memory Ontology Skill

概述

利用 memory MCP server 维护一个结构化知识图谱,跨对话持久记住:

  • 人员(开发者、用户、联系人)
  • 项目(代码库、功能模块、任务)
  • 实体关系(谁负责什么,什么依赖什么)

依赖

  • MCP 服务器:
    memory
    (已配置于
    .vscode/mcp.json

工作流程

存储记忆

  1. 识别对话中提到的关键实体(人、项目、配置、决策)
  2. 调用
    memory/create_entities
    创建实体节点
  3. 调用
    memory/create_relations
    建立实体关联
  4. 调用
    memory/add_observations
    添加观察笔记

召回记忆

  1. 用户提问时,先调用
    memory/search_nodes
    检索相关记忆
  2. 将检索到的上下文注入回复
  3. 更新已有实体的观察记录

标准实体类型

Person:   {name, role, contact, expertise[]}
Project:  {name, path, tech_stack[], status, owner}
Task:     {title, priority, status, assignee, due_date}
Decision: {description, rationale, date, affected_modules[]}
Algorithm:{name, paper_ref, status, weight_in_fusion}

使用示例

用户: 记住张三负责前端,李四负责算法引擎
→ 创建 Person(张三, role=frontend), Person(李四, role=algorithm)
→ 创建关系: 张三-OWNS-kl8-frontend, 李四-OWNS-backend/src/core

用户: KL8 系统中哪个引擎权重最高?
→ 搜索 Algorithm entities, 返回 BayesianEngine(0.20), TransferEntropyEngine(0.18)

自动触发

  • 每次对话完成一个重要决策时,自动调用存储
  • 每次用户说"记住..."、"记录一下..."时立即触发