← 返回 Skills 市场
124
总下载
0
收藏
0
当前安装
5
版本数
在 OpenClaw 中安装
/install xiaodi-obsidian-brain-pro
功能描述
WhatsApp/Telegram → Obsidian 智能记忆增强,防幻觉纠偏,隐私脱敏,让碎碎念自动变成结构化知识(开车/工地语音输入友好)
安全使用建议
Before installing or enabling this skill:
- Review and test scripts in a sandbox: inspect push_to_git.sh/sync.sh and search_engine.py and run them manually with test data to see what is pushed and where.
- Configure OLLAMA_BASE_URL explicitly (prefer localhost) — do not rely on the default http://192.168.3.120:11434.
- Disable or verify Git backup before first run: set git_backup.enabled to false in _meta.json or ensure the repo/remote is a private test repo. Confirm what remote 'origin' points to and that you control it.
- Do not run with secrets present in your Obsidian vault until you are confident redaction works; the redaction is heuristic and can miss unusual token formats. Consider sanitizing notes before enabling auto-push.
- Limit the agent's permissions during testing (run as a user with minimal Git credentials, or run inside an isolated container/VM with a disposable vault).
- If you plan to use Ollama, deploy it locally and point OLLAMA_BASE_URL to a host you control; confirm embedding requests do not leak to external networks.
- If uncertain, run search_engine.py manually to verify behavior and outputs and verify no unexpected network endpoints are contacted.
These checks will reduce the risk of accidental data leakage or misconfiguration. If you want, I can list the exact lines to change to disable Git push by default or to tighten the redaction heuristics.
功能分析
Type: OpenClaw Skill
Name: xiaodi-obsidian-brain-pro
Version: 1.1.4
The skill bundle provides a legitimate utility for syncing Obsidian notes to a Git repository and performing local semantic search via Ollama. It includes proactive security measures in its system prompts (prompts/organizer.txt), specifically instructing the AI agent to redact sensitive information like API keys and passwords before saving or pushing data. While the search script (scripts/search_engine.py) contains a hardcoded local IP address (192.168.3.120) and the sync scripts (scripts/sync.sh) execute shell commands for Git operations, these behaviors are consistent with the stated purpose of the tool and show no evidence of malicious intent or unauthorized data exfiltration.
能力评估
Purpose & Capability
Name/description (WhatsApp/Telegram → Obsidian, local Ollama embeddings, optional Git backup) align with the provided scripts and prompts: the package reads/writes Obsidian notes, runs a local Ollama embedding call, formats notes, and pushes to Git. Minor oddity: no declared required env vars but the Python script uses OLLAMA_BASE_URL (with a non-local default) and OBISIDIAN paths are set in config; overall capability list is coherent with files and metadata.
Instruction Scope
Runtime instructions and scripts will read notebooks from user-specified Obsidian folders, transform content, run a search engine against those files, and execute Git push operations. The skill's policies claim automatic 'privacy redaction' before push, but redaction is implemented as heuristics in prompts (length-based API key detection, token prefixes like ghp_, sk-) which can be bypassed by different token formats or poorly parsed content. Because the skill will read arbitrary .md files and may push them to a remote repo, this scope includes potential data exfiltration if a remote is misconfigured or redaction fails.
Install Mechanism
No external install spec or remote downloads are present — the skill is instruction+scripts only. That lowers supply-chain risk. The included scripts and Python code are plain-text and execute locally; nothing is fetched from third-party URLs during install.
Credentials
The skill declares no required environment variables but the search_engine.py expects OLLAMA_BASE_URL and OLLAMA_MODEL and uses a default OLLAMA_BASE_URL of http://192.168.3.120:11434. That embedded default IP is unexpected and could cause network calls to an unintended host if the user doesn't configure the env var. Git push scripts make network connections to whatever Git remote is configured; secrets required for Git (SSH keys or tokens) are not requested explicitly but will be used if present on the host. The redaction rules are heuristic and may not reliably catch all sensitive tokens, so requesting no credentials reduces transparency rather than risk — the skill still requires network access and may leverage host-level Git credentials.
Persistence & Privilege
The skill is not 'always: true' and does not request modification of other skills. It does require file read/write and exec permissions (declared in metadata) to operate on your Obsidian vault and run git scripts — this is consistent with its purpose but is a persistent local presence with the ability to read, write, and push data from configured folders.
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 xiaodi-obsidian-brain-pro - 安装完成后,直接呼叫该 Skill 的名称或使用
/xiaodi-obsidian-brain-pro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
v1.1.4
更新 README 自述文件
v1.1.3
添加12个关键词tags提升搜索可见度: obsidian/memory/note-taking/voice-input/privacy/ollama...
v1.1.2
SKILL.md一致性修复: homepage URL/安装命令/版本号同步
v1.1.1
添加路径配置警告,提醒用户修改 obsidian_path
v1.1.0
v1.1.0 adds enhanced privacy and smarter input correction:
- Added multi-language correction whitelist for common speech recognition errors (e.g., "Open Crow" → "OpenClaw").
- Introduced automatic privacy redaction for sensitive content such as API keys, passwords, tokens, and phone numbers.
- Improved user experience for WhatsApp/Telegram to Obsidian workflows, ensuring raw user statements are retained and protected.
元数据
常见问题
Xiaodi Obsidian Brain Pro 是什么?
WhatsApp/Telegram → Obsidian 智能记忆增强,防幻觉纠偏,隐私脱敏,让碎碎念自动变成结构化知识(开车/工地语音输入友好). 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 124 次。
如何安装 Xiaodi Obsidian Brain Pro?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install xiaodi-obsidian-brain-pro」即可一键安装,无需额外配置。
Xiaodi Obsidian Brain Pro 是免费的吗?
是的,Xiaodi Obsidian Brain Pro 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Xiaodi Obsidian Brain Pro 支持哪些平台?
Xiaodi Obsidian Brain Pro 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 Xiaodi Obsidian Brain Pro?
由 mx6315909(@mx6315909)开发并维护,当前版本 v1.1.4。
推荐 Skills