← 返回 Skills 市场
Ultramemory
作者
jared-goering
· GitHub ↗
· v0.2.4
· MIT-0
85
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install ultramemory
功能描述
Structured AI agent memory with temporal versioning, relational tracking, and semantic search. Use when storing facts, recalling context, searching past conv...
安全使用建议
This skill behaves like a local structured memory and legitimately needs an LLM API key to perform 'ingest' operations. Before installing or providing credentials: 1) Note the registry metadata omits required env vars even though SKILL.md requires ANTHROPIC_API_KEY or OPENAI_API_KEY — treat that as a transparency issue. 2) Review the upstream ultramemory PyPI/GitHub project (code and issues) because pip install runs code from that package and it will have access to any API keys you set. 3) Consider creating a limited/quota-restricted LLM API key for this use and point ULTRAMEMORY_DB to a directory you control. 4) If you plan to use the optional API server, run it on localhost only and verify firewall/host settings. 5) If you want higher assurance, run the package in an isolated environment (container or VM) and audit network calls during ingest to ensure data doesn’t leave your intended endpoints.
能力评估
Purpose & Capability
The name/description (structured memory, semantic search, temporal versioning) match the code and SKILL.md: the CLI wrappers and startup recall target a local SQLite DB and a local API server. However the registry metadata lists no required environment variables or credentials while SKILL.md and the scripts clearly require an LLM API key (ANTHROPIC_API_KEY or OPENAI_API_KEY). That metadata mismatch is an inconsistency to be aware of.
Instruction Scope
Runtime instructions and scripts stay within the declared purpose: ingest/search/recall against a local SQLite DB, build embeddings (local model), run an optional local API server, and call a localhost endpoint at startup. The ingest path requires an LLM API for fact extraction; there are no instructions to read arbitrary host files or phone home to unexpected external endpoints (network calls shown are to PyPI/GitHub for installation and to http://localhost:8642 for the API).
Install Mechanism
There is no formal install spec in the registry, but SKILL.md instructs pip install ultramemory (PyPI) or git clone from GitHub. Installing a PyPI package is normal, but it grants code execution during install/runtime. The SKILL.md also mentions downloading a local embedding model (~80MB) on first run. The lack of an explicit install record in the registry metadata (despite code files relying on an installed package) is an inconsistency to check.
Credentials
The environment requirements shown in SKILL.md are minimal and appropriate for the feature set (an LLM API key for fact extraction and an optional DB path). However the registry metadata claims no required env vars while the runtime clearly requires ANTHROPIC_API_KEY or OPENAI_API_KEY for ingest. That mismatch reduces transparency. Also note: providing an LLM API key to a third-party package gives that code the ability to make API calls using your credentials — audit the ultramemory package/source before handing over keys.
Persistence & Privilege
The skill does not request global 'always' inclusion or modify other skills or system-wide agent settings. It uses a local DB path and optional local API server. No elevated runtime privileges or persistent agent-wide modifications are requested by the included scripts.
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 ultramemory - 安装完成后，直接呼叫该 Skill 的名称或使用
/ultramemory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
v0.2.4
Security hardening: removed sys.path injection from configurable dirs, removed auto_ingest scripts that read local files, require pip install only. Startup-recall and memory CLI remain.
v0.2.3
Fix security scan: removed auth-profiles.json references and hardcoded paths. Declared LLM API key requirement in metadata. All scripts now use environment variables only.
v0.2.2
Initial ClawHub release. Structured agent memory with atomic fact extraction, relational versioning, semantic search, and three-layer architecture (MEMORY.md + plugin + direct API).
元数据
常见问题
Ultramemory 是什么？
Structured AI agent memory with temporal versioning, relational tracking, and semantic search. Use when storing facts, recalling context, searching past conv... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 85 次。
如何安装 Ultramemory？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install ultramemory」即可一键安装，无需额外配置。
Ultramemory 是免费的吗？
是的，Ultramemory 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Ultramemory 支持哪些平台？
Ultramemory 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 Ultramemory？
由 jared-goering（@jared-goering）开发并维护，当前版本 v0.2.4。
推荐 Skills