Token导航 LogoToken导航TokenDH.com
效率只读clawhub未标认证来源可访问clear审计提醒

rememberremember 搜索

Agent Skill

remember 用于补充效率相关能力,适合在 OpenClaw 中需要让 Agent 承接效率相关任务时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

46,523

周安装

1,978

GitHub Stars

2

下载量

16,299
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:remember(remember 搜索)
来源仓库:https://github.com/ivangdavila/remember
安装命令:
openclaw skills install remember
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install remember

简介

remember 管理持久记忆,自动过滤重要信息并按功能组织长期知识。

  • 适用于需要跨会话保持上下文或积累经验的学习型 Agent。
  • 支持衰减机制,优先保留高频高价值内容,减少冗余干扰。
  • 安装命令:openclaw skills install remember;需配置存储后端如本地文件或云数据库。
  • 建议定期清理过期记忆,防止知识陈旧影响推理准确性。

SKILL.md

name
Remember
description
Curate persistent memory that actually helps. Filter what matters, organize by function, decay what doesn't.
version
1.1.0

The Problem with Most Memory

Storing everything creates noise. Wrong retrieval is worse than no memory. The goal isn't maximum recall — it's retrieving the right thing at the right time.

What's Actually Worth Remembering

High value (persist indefinitely):

  • Commitments made — "I said I'd do X by Y"
  • Learned corrections — "User told me NOT to do Z"
  • Explicit preferences — "I hate verbose responses"
  • Core relationships — "Maria is the designer on Project X"

Medium value (persist with review):

  • Project/context state — what's active, current status
  • Domain lessons — patterns, gotchas, how things work here
  • Decisions made — what was chosen and why

Low value (don't persist):

  • One-off questions, easily reconstructible facts, transient context

Organize by Function, Not Content

Structure by how you'll retrieve it. Adapt categories to your domain:

memory/
├── commitments.md    # Promises, deadlines (with dates!)
├── preferences.md    # Likes/dislikes, style, boundaries
├── corrections.md    # Mistakes not to repeat
├── decisions.md      # What was decided and why
├── relationships.md  # People, roles, context
└── contexts/         # Per-project or per-client state
    └── {name}.md

Memory Hygiene

Every entry needs:

  • Date recorded (when did I learn this?)
  • Source hint (explicit statement vs inference)
  • Confidence (certain / likely / guess)

Prune aggressively:

  • Completed commitments older than 30 days → archive
  • Inactive contexts → move to archive/
  • Contradictions → keep newest, note the change

The staleness test: "If retrieved in 6 months, will this help or mislead?"

Handling Contradictions

When new info conflicts with old:

  1. Don't silently overwrite — note the change
  2. Keep the newer version as active
  3. Optionally log: [Updated 2026-02-11] Was: X, Now: Y

User Control

  • "Remember this" → explicit save with category
  • "Forget that" → explicit delete
  • "What do you know about X?" → transparency
  • "Never remember Y" → hard privacy boundary

See categories.md for domain-specific templates. See consolidation.md for the review/prune process.


*Related: reflection (self-evaluation), loop (iterative refinement)*

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

75.31%
按下载量换算12,275

安全审计

VirusTotal

可疑

ClawScan

通过

Static analysis

未展示

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills