← 返回 Skills 市场
334
总下载
0
收藏
0
当前安装
16
版本数
在 OpenClaw 中安装
/install second-brain-ai
功能描述
Read, capture, search, relate, and assemble context from a user-specified local Markdown knowledge base (Obsidian/Logseq style). Supports controlled write op...
使用说明 (SKILL.md)
Second Brain AI Skill v2.0 (Repair Build)
A lightweight skill for working with a user-chosen Markdown knowledge base with controlled write operations and attribution requirements.
Requirements
- Node.js >= 16.0.0
- Environment variable
SECOND_BRAIN_VAULTmust be set explicitly - Optional: Frontmatter support (YAML)
- Optional: WikiLinks support
[[Note Title]]
Configuration
export SECOND_BRAIN_VAULT="/absolute/path/to/your/vault"
Safety Boundaries
- Only operates within the configured vault path
- Write operations require
allow_write: true - Append operations require
appended_byattribution
Tools
1. init_vault
Initialize a new vault with standard folder structure.
Input: { "allow_write": true }
2. capture_note
Create a new note.
Input:
{
"allow_write": true,
"title": "Note Title",
"content": "Body content",
"type": "idea",
"tags": ["tag1", "tag2"],
"links": ["Related Note"]
}
3. append_note
Append content to an existing note with attribution.
Input:
{
"allow_write": true,
"title": "Note Title",
"content": "Additional content",
"section": "Updates",
"appended_by": "Agent Name"
}
Required: appended_by must identify who is appending.
4. search_notes
Search notes by keywords.
Input: { "query": "search terms", "limit": 5 }
5. find_related
Find notes related to a topic.
Input: { "topic": "Topic Name", "limit": 5 }
6. get_backlinks
Get notes that link to a specific note.
Input: { "note_title": "Target Note" }
7. build_context_pack
Build a context pack for agent consumption.
Input: { "topic": "Topic", "limit": 10 }
8. suggest_links
Get smart link suggestions for a note.
Input: { "title": "Note Title", "limit": 5 }
9. rebuild_index
Refresh index (currently returns skip status as SQLite is not implemented).
Input: {}
Note Format
Standard frontmatter:
---
id: 20260313
title: Note Title
type: idea
tags: [tag1, tag2]
created: 2026-03-13
updated: 2026-03-13
status: active
---
Append Attribution Format
When appending, the skill adds:
> Append Record
> Added by: {appended_by}
> Added at: {timestamp}
{content}
Limitations (Repair Build)
- SQLite indexing is not implemented; uses file-based scanning
- No semantic/vector search
- No automatic link insertion
- No conflict detection
- Single vault only
Testing
npm test
安全使用建议
This skill appears to do what it claims and has no network traffic or secret-collection behavior in the code. Before installing or enabling write operations: 1) Set SECOND_BRAIN_VAULT to a dedicated vault directory (not /, your home directory, or other sensitive folders) so the skill's file access is limited to intended notes; 2) Only enable write actions (pass allow_write=true) when you explicitly approve changes, and ensure appended_by is provided for attribution; 3) Back up your vault before running write operations (init, capture, append) in case of accidental changes; 4) Note the minor metadata mismatch: registry said no required env var but package.json and SKILL.md require SECOND_BRAIN_VAULT — ensure the platform will allow you to set that environment variable. If you need higher assurance, review the included scripts locally or run them in a sandboxed environment before granting write permission.
功能分析
Type: OpenClaw Skill
Name: second-brain-ai
Version: 2.3.1
The second-brain-ai skill bundle is a legitimate tool for managing local Markdown knowledge bases (e.g., Obsidian). It implements functional logic for searching, linking, and creating notes with safety features such as explicit write-approval flags (allow_write) and filename sanitization to prevent path traversal. The code relies solely on Node.js built-in modules (fs, path), contains no external dependencies in package.json, and lacks any network-reaching or shell-executing capabilities, ensuring all operations remain local to the user-defined vault path.
能力评估
Purpose & Capability
Name/description, SKILL.md, package.json, and the included scripts all align: they read and (with explicit flags) write Markdown files inside a user-specified vault. Required runtime (Node >=16 and SECOND_BRAIN_VAULT) is appropriate for this purpose. The only minor inconsistency is that the registry metadata reported 'Required env vars: none' while package.json and SKILL.md require SECOND_BRAIN_VAULT.
Instruction Scope
The SKILL.md instructions map directly to the scripts' behavior. Scripts operate only on files under VAULT_PATH, and write operations require an explicit allow_write=true and, for appends, an appended_by attribution. The instructions do not direct the agent to read other system files or call external endpoints.
Install Mechanism
There is no install spec (instruction-only install), and the repository includes the Node scripts directly. No remote downloads, package installs, or brew/npm installs are performed by the skill itself. Risk here is low because the skill runs included JavaScript files locally.
Credentials
Only one env var (SECOND_BRAIN_VAULT) is required and is non-secret per package.json. This is proportional to the stated purpose. Caveat: the skill's authority equals whatever path the user supplies — if a user sets VAULT to a broad or sensitive directory (e.g., /, /home, or a parent of other data), the skill will be able to read (and, if allow_write is set, write) any .md files there. The skill does not request unrelated credentials or secrets.
Persistence & Privilege
The skill is not always:true and does not claim elevated system privileges. It does not alter other skills or system-wide agent settings. It writes only to the configured vault path and requires explicit allow_write for modifications.
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 second-brain-ai - 安装完成后，直接呼叫该 Skill 的名称或使用
/second-brain-ai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
v2.3.1
V2.3.1 metadata fix: explicitly declare required SECOND_BRAIN_VAULT environment variable to match SKILL.md and runtime behavior.
v2.3.0
V2.3.0 security republish: removed flagged child_process-based test artifact, kept controlled write and append attribution, refreshed release packaging.
v2.0.0
v2.0.0 repair release: mandatory append attribution, repaired scripts, test suite passing, clarified boundaries.
v2.1.7
Remove child_process test runner from published package.
v2.1.6
Resolve vault_path per call and remove module-level path reliance.
v2.1.5
Prefer explicit vault_path input and make env var optional fallback.
v2.1.4
Declare required SECOND_BRAIN_VAULT in visible release metadata.
v2.1.3
Remove published test runner shell execution pattern.
v2.1.2
Align instructions and metadata with controlled-write behavior.
v2.1.1
Released from the v1.0.0 branch line with controlled-write safety fixes.
v2.1.0
Controlled-write release: restore core write actions with explicit allow_write confirmation and strict vault scoping.
v2.0.5
Security remediation: convert to read-only mode and remove mutation scripts.
v2.0.4
Security remediation: remove native dependency and align instructions with shipped behavior.
v2.0.2
Re-publish from the security-hardened codebase aligned to the GitHub v1.0.0 release baseline.
v2.0.1
Security hardening: require explicit vault path, remove default home-directory target, add missing dependency, and publish missing source files.
v1.0.0
Initial release.
元数据
常见问题
Second Brain AI 是什么？
Read, capture, search, relate, and assemble context from a user-specified local Markdown knowledge base (Obsidian/Logseq style). Supports controlled write op... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 334 次。
如何安装 Second Brain AI？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install second-brain-ai」即可一键安装，无需额外配置。
Second Brain AI 是免费的吗？
是的，Second Brain AI 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Second Brain AI 支持哪些平台？
Second Brain AI 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 Second Brain AI？
由 haidong（@harrylabsj）开发并维护，当前版本 v2.3.1。
推荐 Skills