AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Memory Manager

skill-cloud99277-kitclaw-memory-manager · by cloud99277

Manage cross-agent persistent memory using a three-layer model (Identity/Session/Knowledge). Use when you need to search past decisions, actions, or learnings across conversations, or when you need to persist important information for future sessions. 当用户提到"记忆""memory""whiteboard""历史决策""跨会话""保存决策""查找历史"时触发。Prefer this for cross-session memory persistence; use brain-link or conversation-distiller…

No reviews yet
0 installs
36 views
0.0% view→install

Install

$ agentstack add skill-cloud99277-kitclaw-memory-manager

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-cloud99277-kitclaw-memory-manager)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
5mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Memory Manager? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

memory-manager

跨 Agent 持久记忆管理 Skill,基于三层记忆模型(身份层 / 会话层 / 知识层)。

快速开始

检索记忆

# 全层检索
python3 ~/.ai-skills/memory-manager/scripts/memory-search.py "关键词"

# 仅检索 L2 会话层(Whiteboard)
python3 ~/.ai-skills/memory-manager/scripts/memory-search.py "关键词" --layer=L2

# 按项目过滤
python3 ~/.ai-skills/memory-manager/scripts/memory-search.py "关键词" --project=agent-os

# 按领域搜索(仅限 L3,覆盖 config.json)
python3 ~/.ai-skills/memory-manager/scripts/memory-search.py "关键词" --scope=dev       # 研发领域
python3 ~/.ai-skills/memory-manager/scripts/memory-search.py "关键词" --scope=content   # 内容/社交
python3 ~/.ai-skills/memory-manager/scripts/memory-search.py "关键词" --scope=personal  # 个人领域

# 组合使用
python3 ~/.ai-skills/memory-manager/scripts/memory-search.py "关键词" --scope=dev --layer=L3 --json

# 输出 JSON(供其他 skill 机器消费)
python3 ~/.ai-skills/memory-manager/scripts/memory-search.py "关键词" --json

更新记忆

# 从文字中直接写入
python3 ~/.ai-skills/memory-manager/scripts/memory-update.py --from-text "决定使用 JSON 而非 chromadb" --type=decision --project=agent-os

# 从文件中提取(Agent-guided:脚本展示文件内容和提取指令,由 Agent 逐条调用 --from-text 写入)
python3 ~/.ai-skills/memory-manager/scripts/memory-update.py --from-file conversation.md --project=agent-os

# 查看当前 Whiteboard
python3 ~/.ai-skills/memory-manager/scripts/memory-update.py --list

如果你要更方便地把当前任务/对话提炼成 1-3 条 L2 条目,优先用:

python3 ../l2-capture/scripts/l2_capture.py \
  --project agent-toolchain \
  --from-text "[decision] 共享稳定知识统一落到 20_Knowledge_Base" \
  --apply

memory-update.py 仍然是底层入口;l2-capture 是更适合日常写入的上层封装。

监听知识库目录

# 安装稳定版 systemd 用户服务(推荐)
bash ~/.ai-skills/memory-manager/scripts/install-knowledge-watch-service.sh

# 启动目录监听
bash ~/.ai-skills/memory-manager/scripts/start-knowledge-watch.sh

# 查看状态
bash ~/.ai-skills/memory-manager/scripts/status-knowledge-watch.sh

# 停止监听
bash ~/.ai-skills/memory-manager/scripts/stop-knowledge-watch.sh

# 卸载稳定版服务
bash ~/.ai-skills/memory-manager/scripts/uninstall-knowledge-watch-service.sh

默认监听目录:

~/knowledge-base (configure in ~/.ai-memory/config.json)

稳定版行为:

  • 使用 systemd --user 托管,登录后自动拉起
  • 失败自动重启
  • start/stop/status 脚本优先走 systemd 服务
  • 发现 .md 新增 / 修改 / 删除后,等待短暂 debounce
  • 自动补全缺失 frontmatter
  • 自动执行 knowledge-search 增量索引
  • 文件日志写入 ~/.ai-memory/logs/obsidian-l3-watch.log
  • Journal 日志可用:journalctl --user -u obsidian-l3-watch.service -n 50 --no-pager

三层记忆模型

| 层级 | 内容 | 存储 | 检索策略 | |------|------|------|---------| | L1 身份层 | 人格、用户画像、行为规则 | CLAUDE.md / 各 Agent 配置 | 始终加载 | | L2 会话层 | Decisions / Actions / Learnings | ~/.ai-memory/whiteboard.json | 按需检索 | | L3 知识层 | 笔记、研究素材、项目资料 | Obsidian / 本地 Markdown | grep 检索 |

> L3 路径在 ~/.ai-memory/config.json 中配置。

什么是 Whiteboard Memory

L2 的核心机制,从对话中提取三类条目:

  • Decision(决策):两个以上方案之间做了明确选择
  • Action(行动):承诺要做但尚未完成的任务
  • Learning(学习):实施中发现的规律或反模式

提取指南见 references/whiteboard-template.md

数据存储

  • L2 数据目录~/.ai-memory/
  • Whiteboard~/.ai-memory/whiteboard.json
  • 配置文件~/.ai-memory/config.json(配置 L3 路径)

首次运行时自动初始化,无需手动创建目录。

设计约束

  • 零外部依赖(纯 Python stdlib)
  • 不引入向量数据库,用 grep 覆盖 80% 检索场景
  • L2/L3 数据与 skill 仓库完全分离(存放在 ~/.ai-memory/
  • 所有数据格式含 schema_version 字段(对标 Phase 1 IO 契约约束)

参考文档

  • references/memory-architecture.md — 三层模型详细设计
  • references/whiteboard-template.md — Whiteboard 提取指南

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.