← 返回 Skills 市场
345
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install hippocampus-subagent-memory
功能描述
Isolate and coordinate sub-agent memory in OpenClaw with Hippocampus using scoped IDs, bounded merge-back, and explicit cross-agent imports.
使用说明 (SKILL.md)
Hippocampus Subagent Memory
Use this skill when OpenClaw spawns sub-agents that need their own memory without contaminating the parent agent context.
Use It For
- creating isolated sub-agent memory namespaces
- preventing accidental memory leakage across siblings
- returning only bounded summaries or artifacts to the parent
- coordinating explicit import from child to parent
Preferred Flow
- Spawn a child with a scoped
HIPOKAMP_SUBAGENT_ID. - Let the child write only into its own namespace.
- Return a bounded result package at the end of the task.
- Import back to parent only what is explicitly approved.
Guidance
- Default to isolation.
- Do not flatten full child memory into the parent.
- Prefer summary, artifacts, and references over transcript copy.
- Tag child memory with session and parent linkage metadata.
- Root-agent bootstrap should happen once in the portal; child agents should inherit scope automatically and should not require separate portal signup.
Related
hippocampus-memory-corefor core memory operationshippocampus-openclaw-onboardingfor base config@hippocampus/openclaw-context-enginefor automated spawn/end lifecycle hooks
安全使用建议
This skill is mostly a policy document, but there are a few things to check before you use it in production: 1) Confirm the intended environment variable name — the doc uses 'HIPOKAMP_SUBAGENT_ID' which looks like a typo; ask the author what the canonical scoped-ID is and how it's set. 2) Ask how the 'portal' bootstrap and related packages are provided: do you need hippocampus-memory-core or onboarding code installed elsewhere? If so, request a concrete install/auth plan and minimal required credentials. 3) If you will let agents inherit scope automatically, verify this does not grant them access to unrelated credentials or agent configuration. 4) Test this in a sandbox: ensure child agents only write to their namespace and that the merge-back mechanism only returns the bounded summaries you expect. 5) Prefer the author add explicit env/credential requirements (or confirm none are needed) and fix the typo to avoid accidental misconfiguration. If those clarifications are provided and implemented minimally, the skill can be considered coherent; without them, treat it cautiously.
功能分析
Type: OpenClaw Skill
Name: hippocampus-subagent-memory
Version: 0.1.0
The skill bundle contains metadata and architectural instructions for managing sub-agent memory isolation. The SKILL.md file provides logical guidance on using scoped IDs and bounded data merging to prevent context contamination between parent and child agents, with no evidence of malicious code, data exfiltration, or harmful prompt injection.
能力评估
Purpose & Capability
Skill name/description (Hippocampus sub-agent memory) matches the instructions (create isolated namespaces, bounded merge-back). However the SKILL.md references related packages and onboarding artifacts (hippocampus-memory-core, hippocampus-openclaw-onboarding, @hippocampus/openclaw-context-engine) and a 'portal' process without declaring any install steps or required credentials—this is plausible for an instruction-only policy but is an unexplained dependency surface. Also the SKILL.md uses a differently spelled env token 'HIPOKAMP_SUBAGENT_ID' which is likely a typo and could cause misconfiguration.
Instruction Scope
The runtime instructions stay within the stated purpose: spawn child agents, restrict writes to a scoped namespace, return bounded summaries, and import explicitly. There are no directives to read arbitrary host files, exfiltrate data, or contact external endpoints outside the conceptual 'portal' mention.
Install Mechanism
This is an instruction-only skill with no install spec and no code files — low disk/write risk. Because no install occurs, there is no direct supply-chain risk from this repository itself.
Credentials
The skill declares no required env vars or credentials, which is proportionate for an instruction doc. But it references a scoped ID value ('HIPOKAMP_SUBAGENT_ID') and a portal bootstrap process without specifying how scope/authentication is obtained. That inconsistency (missing declared env usage, and the probable typo) should be clarified; otherwise implementers may hardcode sensitive tokens or invent unsafe workarounds.
Persistence & Privilege
The skill does not request permanent presence (always:false) and has no install steps that modify agent/system configuration. Autonomous invocation is allowed (platform default) but not combined here with other clear red flags.
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 hippocampus-subagent-memory - 安装完成后，直接呼叫该 Skill 的名称或使用
/hippocampus-subagent-memory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
v0.1.0
- Initial release of hippocampus-subagent-memory skill.
- Enables isolated sub-agent memory in OpenClaw using scoped IDs.
- Prevents accidental memory leakage across sibling/parent contexts.
- Supports bounded merge-back and explicit import from child to parent.
- Includes guidance for namespace management and session linkage.
元数据
常见问题
Hippocampus Subagent Memory 是什么？
Isolate and coordinate sub-agent memory in OpenClaw with Hippocampus using scoped IDs, bounded merge-back, and explicit cross-agent imports. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 345 次。
如何安装 Hippocampus Subagent Memory？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install hippocampus-subagent-memory」即可一键安装，无需额外配置。
Hippocampus Subagent Memory 是免费的吗？
是的，Hippocampus Subagent Memory 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Hippocampus Subagent Memory 支持哪些平台？
Hippocampus Subagent Memory 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 Hippocampus Subagent Memory？
由 cezexPL（@cezexpl）开发并维护，当前版本 v0.1.0。
推荐 Skills