← 返回 Skills 市场
ruiyongwang

Oclaw Hermes

作者 ruiyongwang · GitHub ↗ · v3.0.0 · MIT-0
cross-platform ⚠ suspicious
115
总下载
0
收藏
1
当前安装
2
版本数
在 OpenClaw 中安装
/install oclaw-hermes
功能描述
OpenClaw × Hermes × DeerFlow 三位一体智能体桥接方案 - 实现 mflow 记忆流同步、智能体集群协作、深度研究链和专家蒸馏。让 OpenClaw、Hermes、DeerFlow 的边界,成为你能力的延伸。
安全使用建议
This package appears to be a substantial bridge/orchestration project that will create local databases, write files under your home directory, and run Docker containers that connect to local and remote services. However, the registry metadata incorrectly claims no required environment variables or config paths while the SKILL.md, docker-compose.yml, and scripts clearly require multiple API keys and will persist memory data. Before installing: 1) do not supply high-privilege tokens (e.g., full cloud credentials); create least-privilege/test tokens instead; 2) review the .env.example and every script for endpoints and persistence locations (~/.openclaw, ~/.hermes, ~/.oclaw-hermes); 3) inspect the Docker images (nousresearch/hermes, bytedance/deerflow) you will pull—prefer pinned image digests and official images; 4) run in an isolated environment or VM first (to avoid leaking secrets or contaminating your real home config); 5) confirm the repository source and maintainership (the registry lists source unknown but SKILL.md references a GitHub repo); and 6) if you need this functionality, ask the author to correct the registry metadata (declare required env vars and config paths) and to provide a minimal test mode that doesn't auto-persist or auto-publish. The mismatches raise enough concern to pause installation until you verify secrets, image provenance, and the exact runtime behavior.
功能分析
Type: OpenClaw Skill Name: oclaw-hermes Version: 3.0.0 The oclaw-hermes bundle is a comprehensive integration framework designed to bridge OpenClaw with the Hermes memory system and DeerFlow multi-agent orchestration. Its core functionality involves a 'Memory Flow' (mflow) engine that captures, categorizes, and synchronizes conversation history and 'expert' insights across platforms using local SQLite storage and configurable API endpoints. While the scripts utilize high-privilege capabilities such as file system access (writing to ~/.oclaw-hermes), network requests (syncing to local/remote services), and shell execution (checking service status via subprocess), these actions are transparently documented and strictly serve the stated purpose of the architecture. The presence of hardcoded Windows file paths in some scripts suggests a specific development environment rather than malicious intent.
能力标签
crypto
能力评估
Purpose & Capability
The skill claims to bridge OpenClaw, Hermes, and DeerFlow (expected to need local services, containers, and tokens), but the registry metadata lists no required environment variables or config paths. In reality SKILL.md, docker-compose.yml, and multiple scripts reference and require tokens/URLs (OPENCLAW_TOKEN, OPENROUTER_API_KEY, ANTHROPIC_API_KEY, DEERFLOW_*), and write to home config directories (~/.openclaw, ~/.hermes, ~/.oclaw-hermes). The omission in metadata is an incoherence: either the metadata is incomplete or the skill is hiding required privileges.
Instruction Scope
SKILL.md explicitly instructs cloning a GitHub repo, creating a .env with API keys, running docker-compose to pull/run multiple containers, and running Python scripts. These instructions will create persistent DBs and memory files and may push data to OpenClaw/Hermes/DeerFlow endpoints. The actions (running containers, creating files under user home, requiring tokens) are coherent for a bridge/orchestrator but grant broad local persistence and network access; the SKILL.md also references commands (e.g., python scripts/verify.py) not present in the manifest, which is a minor inconsistency.
Install Mechanism
No formal install spec in the registry (instruction-only), but the project includes docker-compose that will pull images (nousresearch/hermes, bytedance/deerflow) and build local Dockerfiles. There are no obscure download URLs in the package itself, but running docker-compose will fetch external container images—this is expected for this type of project but increases risk surface (third-party images, network pulls).
Credentials
Registry metadata claims no required env vars, yet SKILL.md, docker-compose.yml, and scripts require multiple credentials and service URLs (OPENCLAW_TOKEN, OPENROUTER_API_KEY, ANTHROPIC_API_KEY, DEERFLOW_GATEWAY_URL, DEERFLOW_LANGGRAPH_URL, etc.). The code also reads/writes to local config paths and persistent DBs under user home. Requesting multiple unrelated model-provider keys and platform tokens is reasonable for a multi-platform bridge, but the mismatch with declared requirements and lack of explicit justification is a red flag.
Persistence & Privilege
The skill will create persistent artifacts (SQLite DBs in ~/.openclaw/.oclaw-hermes, files under ~/.hermes, Docker containers and volumes) and can sync memories across platforms. It does not set always:true, but it does request persistent local storage and may auto-register/publish skills per config options. Persistence and container orchestration are expected for this use case but increase blast radius; verify storage locations and retention policy before use.
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 oclaw-hermes
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /oclaw-hermes 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
v3.0.0
oclaw-hermes v3.0.0 is a major update with deep system fusion, memory-driven routing, and new unified architecture. - Introduced UnifiedCore to deeply merge OpenClaw, Hermes, and DeerFlow systems. - Added mflow v2.0: five-layer memory flow for real-time, cross-platform memory synchronization. - Implemented memory-driven agent routing and Skill-memory dual enhancement. - Expanded agent cluster to six core roles (Lead/Research/Code/Browser/Memory/Skill). - Enabled automatic memory extraction and graph-based memory reasoning. - Added scripts: auto_memory.py, mflow_v2.py, store_memory.py, and unified_core.py. - Updated documentation to reflect new architecture and features.
v2.0.0
**Major update: oclaw-hermes 2.0.0 integrates OpenClaw, Hermes, and DeerFlow, unlocking seamless multi-agent collaboration and advanced research capabilities.** - Introduces unified memory synchronization (mflow) across OpenClaw, Hermes, and DeerFlow. - Adds multi-agent cluster management with roles like Lead, Research, Code, Browser, Skill, Memory, and Expert Agent. - Supports deep research chains for one-click advanced topic analysis and report generation. - Provides local and Docker deployment options, with detailed configuration for skills, agents, and LLM providers. - Includes command-line and Python API interfaces for messaging, research, expertise distillation, and memory sync. - Enhanced skill ecosystem integration and expert knowledge distillation.
元数据
Slug oclaw-hermes
版本 3.0.0
许可证 MIT-0
累计安装 1
当前安装数 1
历史版本数 2
常见问题

Oclaw Hermes 是什么?

OpenClaw × Hermes × DeerFlow 三位一体智能体桥接方案 - 实现 mflow 记忆流同步、智能体集群协作、深度研究链和专家蒸馏。让 OpenClaw、Hermes、DeerFlow 的边界,成为你能力的延伸。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 115 次。

如何安装 Oclaw Hermes?

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

Oclaw Hermes 是免费的吗?

是的,Oclaw Hermes 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Oclaw Hermes 支持哪些平台?

Oclaw Hermes 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 Oclaw Hermes?

由 ruiyongwang(@ruiyongwang)开发并维护,当前版本 v3.0.0。

💬 留言讨论