← 返回 Skills 市场
Memory Manager
作者
onlyloveher
· GitHub ↗
· v1.0.0
· MIT-0
122
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install memory-manager-clawd
功能描述
Local memory management for agents. Compression detection, auto-snapshots, and semantic search. Use when agents need to detect compression risk before memory...
安全使用建议
This skill appears to implement local memory management as advertised and contains only shell scripts that operate on local files (no network calls). However: 1) Back up your workspace before running—organize.sh and categorize.sh use mv/cp and can move files unexpectedly. 2) Confirm which directory will be used: scripts default to $HOME/.openclaw/workspace (or OPENCLAW_WORKSPACE) but skill.yaml suggests ~/clawd/memory—set OPENCLAW_WORKSPACE to the folder you expect. 3) The package/metadata/homepage/ownerId values are inconsistent across files—treat metadata with caution and verify the source or author if provenance matters. 4) Review the scripts (they're small) in a sandbox or run them with echo/dry-run to confirm behavior; ensure jq is available if you want JSON state updates. If these concerns are acceptable and you keep a backup, the skill is coherent; if provenance or accidental file movement is a concern, do not install or run it until resolved.
功能分析
Type: OpenClaw Skill
Name: memory-manager-clawd
Version: 1.0.0
The skill implements a structured memory management system for agents but contains several shell script vulnerabilities due to a lack of input sanitization. Specifically, 'categorize.sh' is vulnerable to path traversal via the 'NAME' parameter, and 'search.sh' is susceptible to flag injection through the 'QUERY' parameter, which could allow unauthorized file access or manipulation within the agent's environment. While these appear to be unintentional flaws rather than intentional malice, they represent risky capabilities that could be exploited through prompt injection.
能力评估
Purpose & Capability
Name/description match the included scripts: init.sh, detect.sh, organize.sh, snapshot.sh, search.sh, categorize.sh, stats.sh all operate on a local workspace memory directory and implement the described three-tier memory model. However there are metadata/config mismatches: skill.yaml's MEMORY_PATH default (~/clawd/memory) does not match the scripts' actual default workspace ($HOME/.openclaw/workspace via OPENCLAW_WORKSPACE). Package/README/SKILL.md reference different homepages (moltbook/clawhub) and _meta.json ownerId differs from provided registry metadata. These inconsistencies could cause confusion about where data is stored or which configuration is authoritative.
Instruction Scope
Runtime instructions and scripts stay local and do not contact external endpoints. They create/read/move/copy files under the workspace memory directory and use common CLI tools (find, grep, mv, cp, wc, jq if available). Important behaviors to note: organize.sh and categorize.sh move or copy user files (mv/cp), snapshot.sh reads and aggregates recent files into snapshot files, and search.sh suggests a 'memory_get' tool that is not included. Because some scripts perform destructive moves (mv), you should backup your memory directory before running; the instructions themselves recommend a backup but the scripts will act on any files present in the target directories.
Install Mechanism
No install spec or external downloads; this is instruction + shell script based. Nothing will be fetched from network during install. That reduces supply-chain risk, but you still execute local shell scripts on your system when you run them.
Credentials
The skill declares no required env vars, which matches the scripts (they only use OPENCLAW_WORKSPACE optionally). However skill.yaml exposes a MEMORY_PATH config that the scripts do not use, creating a mismatch between declared configuration and actual runtime behavior. There are no secret/credential requests. The scripts will act on whatever path OPENCLAW_WORKSPACE points to (or the default), so verify that path before running.
Persistence & Privilege
The skill is not always-enabled and does not request elevated privileges or modify other skills. It writes state and snapshots to its own workspace directories only.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install memory-manager-clawd - 安装完成后，直接呼叫该 Skill 的名称或使用
/memory-manager-clawd触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
- Initial release of Memory Manager skill.
- Implements a three-tier memory system: episodic, semantic, and procedural memory.
- Provides tools for compression risk detection, auto-snapshots, and organized memory structure.
- Supports quick initialization, context organization, keyword-based search, and manual categorization.
- Includes scripts for stats monitoring, safe migration from flat file structures, and periodic health checks.
元数据
常见问题
Memory Manager 是什么？
Local memory management for agents. Compression detection, auto-snapshots, and semantic search. Use when agents need to detect compression risk before memory... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 122 次。
如何安装 Memory Manager？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install memory-manager-clawd」即可一键安装，无需额外配置。
Memory Manager 是免费的吗？
是的，Memory Manager 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Memory Manager 支持哪些平台？
Memory Manager 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Memory Manager？
由 onlyloveher（@onlyloveher）开发并维护，当前版本 v1.0.0。
推荐 Skills