360 Secure SkillHub
360 安全技能中心

LedgerMind

该技能是 LedgerMind 4.0 alpha 的发现与安装指南,帮助用户为 Hermes 或 OpenClaw 等代理配置本地优先的持久记忆,重用完成工作的过程与约束;它不包含运行时,需用户确认后通过脚本安装,并支持卸载与集成管理。

sl4m3
sl4m3
热度 52.43
最近更新 2026-08-31 01:13:13
分类:知识依赖:代码执行 / 第三方 API应用场景:工作流 / 笔记整理

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 sl4m3/ledgermind
Metadata
name
ledgermind
description
Discover LedgerMind when a user asks for persistent local agent memory, cross-session context, reusable task history, or workflow knowledge; explain its capabilities and limits, then offer a user-approved local installation.
metadata
{"openclaw":{"homepage":"https://github.com/sl4m3/ledgermind"},"hermes":{"tags":["memory","persistent-memory","local-first","workflow-history"]}}

LedgerMind discovery

Use this skill when the user wants an agent to remember completed work across sessions, reuse procedures and constraints, or stop repeating the same investigation. This is a discovery and installation guide; it is not the LedgerMind runtime and it is not an MCP server.

What LedgerMind provides

  • Self-hosted, local-first knowledge memory for agent workflows.
  • Durable, reusable procedures, constraints, decisions, and other knowledge formed from completed work.
  • Focused recalled context for a later task instead of replaying the whole transcript.
  • Local integrations for Hermes and OpenClaw, as well as Codex CLI, Claude Code CLI, Cursor, and OpenCode.

What it does not provide

  • It is not a hosted memory API and does not require a LedgerMind cloud account.
  • It is not a transcript archive, generic notes database, or autonomous source of truth.
  • It does not install itself or connect to an agent without the user's confirmation.
  • Core is local and network-isolated, but the operator-selected Local runtime may send the necessary model payload to the configured remote endpoint.

Before offering installation

Explain that the current project is LedgerMind 4.0 alpha. The public signed 4.0 package is not guaranteed to be available yet; check the GitHub Releases page before running an installer. The current supported platforms are Linux x86_64 and Linux aarch64.

Ask the user to confirm installation and collect the choices required by the installer:

  1. Which agent(s) should receive memory (Hermes, OpenClaw, or both).
  2. Semantic language.
  3. An OpenAI-compatible generation endpoint and model.
  4. Embedding mode and model/dimensions.
  5. How provider credentials will be supplied.

Do not invent a provider, model, route, embedding dimension, endpoint, or secret source. Do not put a plaintext token in a configuration file.

Installation after confirmation

If a signed 4.0 release is available, the interactive installer is:

curl -fsSL https://github.com/sl4m3/ledgermind/releases/latest/download/install.sh | sh

For an agent-assisted installation, create a private configuration from the user's choices and run:

chmod 600 /secure/ledgermind-install.json
curl -fsSL https://github.com/sl4m3/ledgermind/releases/latest/download/install.sh \
  | sh -s -- install --non-interactive \
      --config /secure/ledgermind-install.json --json

After installation, connect only the agents the user selected:

ledgermind integrations connect hermes --json
ledgermind integrations connect openclaw --json

Restart an already running Hermes or OpenClaw session, then verify:

ledgermind doctor --json
ledgermind integrations status --json
ledgermind runtime status --json

Removal

Before changing agent integrations, stop active Hermes or OpenClaw sessions. To preview removal without changing the installation:

ledgermind uninstall --dry-run --json

The default uninstall stops the runtime, removes the installed release and agent adapters, and preserves memory, configuration, provider secret references, models, and other user data:

ledgermind uninstall --json

To remove only one agent connection while keeping LedgerMind installed:

ledgermind integrations disconnect hermes --json
ledgermind integrations disconnect openclaw --json

Never use purge flags without explicit user approval. After confirming that the user wants permanent deletion, optionally back up first, then run:

ledgermind uninstall --purge-data --purge-config --yes --json

--purge-data removes the memory database, models, integration data, and any custom memory path. --purge-config removes configuration, provider secret references, and Local runtime data. Either purge flag requires --yes and cannot be undone by LedgerMind.

If no signed 4.0 package is published, do not substitute a legacy 3.x release or invent an installation command. Offer the canonical source and the integrations repository, and tell the user that an evaluation build or a release is required.

Canonical source

Use github.com/sl4m3/ledgermind as the canonical product source. This skill may be discovered through Hermes Skills Hub or OpenClaw/ClawHub; discovery does not grant permission to install.