360 Secure SkillHub
360 安全技能中心

OpenClaw Memory

围绕本地 OpenClaw 记忆的复用与维护：先查长期与每日 markdown 记忆文件，必要时回退到 SQLite 分块文本，抽取主机、路径、运行手册等具体事实，并按长期规则或当日进展写入对应文件后报告更新路径。

kamiender
kamiender
热度 40.80
最近更新 2026-05-11 17:36:15
分类：知识依赖：代码执行 / 读写文件应用场景：笔记整理 / 工作流 / 研究

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add openclaw-shared-memory
Metadata
name
openclaw-memory
description
Reuse, inspect, search, or append shared OpenClaw memory on the current machine. Use when a request depends on remembered local servers, deploy hosts, internal services, script paths, runbooks, workspace conventions, or prior user preferences that may already exist in OpenClaw memory.

OpenClaw Memory

Resolve the skill directory from this SKILL.md file, then use the bundled script at scripts/openclaw_memory.py.

Data sources

  • Long-term memory: ~/.openclaw/workspace/MEMORY.md
  • Daily memory: ~/.openclaw/workspace/memory/YYYY-MM-DD.md
  • Indexed chunk fallback: ~/.openclaw/memory/main.sqlite

If OPENCLAW_HOME is set, use that as the root instead of ~/.openclaw.

Workflow

  1. If the request names a known machine, service, workflow, or environment-specific noun, search memory before any live checks.
  2. Prefer markdown files as the source of truth when inspecting or updating memory.
  3. If markdown results are sparse, query SQLite chunk text as a fallback.
  4. Extract concrete facts such as hostnames, IPs, ports, file paths, runbook steps, and user preferences.
  5. When the user wants to persist something:
    • Write durable rules and preferences to MEMORY.md.
    • Write day-specific progress to memory/YYYY-MM-DD.md.
  6. After writing, report the exact file that was updated.

Constraints

  • Do not claim semantic or vector retrieval is active unless you actually verified the runtime supports it.
  • For Chinese queries, direct markdown scanning is often more reliable than SQLite tokenization.
  • If the OpenClaw paths are missing, say so plainly and stop instead of inventing state.

Commands

Replace <skill-dir> with the directory that contains this SKILL.md.

python3 <skill-dir>/scripts/openclaw_memory.py doctor
python3 <skill-dir>/scripts/openclaw_memory.py recent --limit 5
python3 <skill-dir>/scripts/openclaw_memory.py search --query "deploy host" --limit 8
python3 <skill-dir>/scripts/openclaw_memory.py append --scope daily --text "Confirmed the deploy host is healthy." --dry-run
python3 <skill-dir>/scripts/openclaw_memory.py append --scope long --text "- Always search both the workspace project tree and the main project tree."

Output expectations

  • For search, return compact results with path, source, score, and a short snippet.
  • For append, print the destination path.
  • For doctor, print which OpenClaw paths exist.