← 返回 Skills 市场
kylin19860916

Structured Vector Memory (SVM)

作者 kylin19860916 · GitHub ↗ · v3.1.0 · MIT-0
cross-platform ⚠ suspicious
287
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install structured-vector-memory
功能描述
Structured Vector Memory (SVM) / 三层高效记忆存储法。整合 LanceDB Pro 向量引擎 + 三层结构化管理,涵盖:每日记忆整理(Micro Sync + Daily Wrapup)、记忆蒸馏压缩(Weekly Compound)、去重检测、scope 隔离、archive 机...
安全使用建议
This skill appears to implement a coherent on-device memory management workflow, but there are notable mismatches you should address before installing: (1) SKILL.md requires 'memory-lancedb-pro' and a Jina API Key for embeddings/rerank, but the registry metadata does not declare any required env vars or credentials — ask the author to declare these explicitly. (2) The installation copies scripts into ~/.openclaw/shared and adds cron jobs that will run automatically and append tasks to HEARTBEAT.md; review those cron entries and the scripts to ensure you're comfortable with scheduled writes to your home workspace. (3) The workflow reads/writes files like AGENTS.md, SOUL.md, USER.md, IDENTITY.md and MEMORY.md which may contain sensitive information; inspect those files and remove or redact sensitive data before enabling automated processes. (4) If you proceed, test in an isolated environment or non-production account, verify how the plugin handles the Jina API Key (where it is stored and whether it's transmitted externally), and request that the skill author update the registry metadata to list required env vars/configs. If you want, ask me to produce a short checklist or a sanitized-install plan (commands to run manually) to reduce risk.
功能分析
Type: OpenClaw Skill Name: structured-vector-memory Version: 3.1.0 The skill bundle implements a structured memory management system (SVM) using LanceDB and a three-tier storage architecture. It utilizes shell scripts (daily-wrapup.sh, micro-sync.sh, weekly-compound.sh) and crontab to automate periodic memory summarization and distillation by appending instructions to a local HEARTBEAT.md file. All operations are confined to the standard OpenClaw workspace, and the logic is transparently documented and consistent with the stated goal of enhancing agent long-term memory without any signs of data exfiltration or malicious intent.
能力评估
Purpose & Capability
The name/description (Structured Vector Memory using LanceDB Pro) align with the provided scripts and instructions for automated memory capture, deduplication, distillation, and archiving. However, SKILL.md lists preconditions (memory-lancedb-pro plugin and a Jina API Key) that are not declared in the registry metadata (no required env vars or primary credential), which is an incoherence.
Instruction Scope
Runtime instructions and included shell scripts read and write multiple files under ~/.openclaw/workspace (MEMORY.md, AGENTS.md, SOUL.md, USER.md, HEARTBEAT.md, archives), append tasks to HEARTBEAT.md, and rely on agent commands like sessions_list/sessions_history/memory_recall/memory_store. These operations are within the stated purpose but involve accessing potentially sensitive identity/agent files and creating cron-driven automation. The scripts append tasks but do not themselves perform the 'recall' or 'store' actions — they assume other agent components will act on HEARTBEAT, which increases implicit privileges and attack surface.
Install Mechanism
There is no network install spec; this is instruction-only with three small shell scripts included. Installation consists of copying scripts into ~/.openclaw/shared and adding crontab entries. No downloads or archive extraction are used.
Credentials
SKILL.md explicitly requires 'memory-lancedb-pro' and a 'Jina API Key' (embedding + rerank), but the registry lists no required env vars or primary credential. The skill also instructs modifying openclaw.json to enable a plugin. The missing declaration of required credentials and plugin configuration is a proportionality/information mismatch and should be fixed before trusting the skill.
Persistence & Privilege
The skill requests persistent scheduled execution via crontab (micro-sync 5×/day, daily wrapup, weekly compound) and writes files under the user's ~/.openclaw workspace. 'always' is false and autonomous invocation is allowed (platform default). Cron-based persistence is expected for this functionality but increases the blast radius (scheduled automated actions that run without interactive confirmation).
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install structured-vector-memory
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /structured-vector-memory 触发
  4. 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
v3.1.0
v3.1: Merged structured-memory (5-layer) into SVM. Entity/relationship management via entity category.
v3.0.0
v3.0: Merged daily-memory + memory-distill into SVM. One skill to rule all memory management.
v2.0.1
Renamed: 三层高效记忆存储法 / Structured Vector Memory (SVM). LanceDB Pro + three-layer structured management.
元数据
Slug structured-vector-memory
版本 3.1.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 3
常见问题

Structured Vector Memory (SVM) 是什么?

Structured Vector Memory (SVM) / 三层高效记忆存储法。整合 LanceDB Pro 向量引擎 + 三层结构化管理,涵盖:每日记忆整理(Micro Sync + Daily Wrapup)、记忆蒸馏压缩(Weekly Compound)、去重检测、scope 隔离、archive 机... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 287 次。

如何安装 Structured Vector Memory (SVM)?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install structured-vector-memory」即可一键安装,无需额外配置。

Structured Vector Memory (SVM) 是免费的吗?

是的,Structured Vector Memory (SVM) 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Structured Vector Memory (SVM) 支持哪些平台?

Structured Vector Memory (SVM) 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 Structured Vector Memory (SVM)?

由 kylin19860916(@kylin19860916)开发并维护,当前版本 v3.1.0。

💬 留言讨论