EO Ability: Memory

主动记忆能力(Proactive Memory)，跨会话延续用户偏好、项目上下文，记住导师要求和格式规范

Security Scan

The skill's declared purpose (proactive, cross-session memory) matches its instructions and it requests no extra privileges or installs — it is internally coherent, but you should confirm where memories are actually stored and who can access them before enabling persistent memory features.

Purpose & Capability

Name/description (proactive memory across sessions) align with the SKILL.md: all examples and the interface focus on save/load/search/forget/update memory operations. The skill does not request unrelated binaries, env vars, or config paths.

Instruction Scope

Runtime instructions and examples limit actions to memory operations and do not instruct the agent to read unrelated files, environment variables, or send data to unexpected endpoints. The SKILL.md references integration with EO/LanceDB but does not include any out-of-scope collection steps.

Install Mechanism

No install spec and no code files — instruction-only skill. This is lowest-risk from an install perspective because nothing is written to disk by the skill itself.

Credentials

The skill declares no required environment variables, credentials, or config paths. It mentions LanceDB/EO integration conceptually but does not request access tokens or other secrets in the SKILL.md.

Persistence & Privilege

always is false (normal). However, the skill's purpose is persistent memory across sessions — that implies storage and access to user data by whatever system implements the memory (EO/LanceDB). Confirm storage location, retention, access controls, and encryption with the EO platform before use.

Assessment

This skill is internally consistent: it only describes memory operations and asks for no extra permissions. Before enabling: 1) Ask where memories will be stored (which database/service and which account). 2) Confirm who can read those memories (other skills, admins, third parties). 3) Check retention and deletion policies and whether data is encrypted at rest/in transit. 4) Test with non-sensitive sample data to verify behavior. If the EO platform requires additional credentials or an external LanceDB endpoint to enable persistence, only provide those after you review that platform's privacy/security controls.

主动记忆能力 (Proactive Memory) - 记住用户偏好、项目上下文，跨会话延续

一句话介绍

跨会话主动记忆能力，记住用户偏好、导师要求、项目上下文，下次对话自动恢复。

核心功能

• 自动记忆: 关键决策自动保存
• 跨会话恢复: 下次对话自动加载相关记忆
• 偏好学习: 根据用户反馈优化策略
• 遗忘机制: 自动清理过期记忆

使用方法

保存记忆

# 自动保存（由 EO 自动触发）
保存: "导师偏好：方法论章节需要更详细的数学推导"
保存: "格式要求：双栏IEEE模板"
保存: "上次讨论焦点：价格谈判"

加载记忆

# 加载用户偏好
加载记忆: "用户偏好"

// 自动恢复
用户: "继续上次的研究"
→ 自动加载:
→   - "目标期刊Medical Image Analysis"
→   - "IEEE格式要求"
→   - "上次模型已经跑出初步结果"

搜索记忆

# 搜索相关记忆
搜索记忆: "关于价格的所有记忆"
搜索记忆: "导师的偏好"

与EO插件的协同

• 被所有 eo-workflow-* 调用
• 被案例2（学术论文研究）使用
• 被案例4（秘书行政工作）重点使用

独立运行模式（有EO vs 无EO）

模式	能力
有EO插件	跨会话持久化、LanceDB索引、语义搜索
无插件（基础）	当前会话记忆、无持久化

示例

// 案例：学术论文
用户: "继续上次的研究，模型已经跑出初步结果"

// EO 自动执行：
const memory = await eo_ability_memory({
  action: 'load',
  context: '当前项目'
})
// → 加载: "用户偏好IEEE格式" + "目标期刊Medical Image Analysis"

// 完成论文撰写后
await eo_ability_memory({
  action: 'save',
  key: '论文进度',
  value: { phase: 'methodology', done: true }
})

Interface

Input

interface MemoryInput {
  action: 'save' | 'load' | 'search' | 'forget' | 'update'
  key?: string
  value?: any
  context?: string
  tags?: string[]
}

Output

interface MemoryOutput {
  success: boolean
  data?: any
  related?: MemoryEntry[]
  memoryUsed?: number
}

---

🦞⚙️ 钢铁龙虾军团