← Back to Skills Marketplace
153
Downloads
0
Stars
0
Active Installs
5
Versions
Install in OpenClaw
/install openclaw-local-embedding
Description
Initialize and configure OpenClaw local embedding mode on CPU-only machines. Handles network probe, proxy fallback, GGUF model download, cmake/llama.cpp comp...
Usage Guidance
The skill appears to do what it says: probe for an HTTP CONNECT proxy, download a HuggingFace GGUF embedding model, and configure OpenClaw. Before running it: (1) review the included script yourself (scripts/init-model.mjs) — it's the active part that will run; (2) be cautious about NODE_TLS_REJECT_UNAUTHORIZED=0 — only use this if you explicitly trust the proxy doing TLS inspection; (3) verify any proxy address you supply or accept (the script tries a hard-coded Kuaishou internal IP which is only appropriate for that cloud); (4) run initial tests in an isolated VM/container if possible, and back up any OpenClaw config before applying changes; (5) ensure openclaw and node-llama-cpp come from sources you trust, and confirm the model is fetched from HuggingFace (the script uses a HuggingFace model id). If you are uncomfortable with disabling TLS verification or using unknown proxies, do not run the script on sensitive systems.
Capability Analysis
Type: OpenClaw Skill
Name: openclaw-local-embedding
Version: 1.0.4
The skill bundle provides a utility for setting up local embeddings but introduces security vulnerabilities by explicitly disabling TLS certificate verification (NODE_TLS_REJECT_UNAUTHORIZED=0) in both SKILL.md and scripts/init-model.mjs. It also includes hardcoded references to an internal proxy IP (10.74.176.8) and performs network probing. While these actions are aligned with the stated purpose of supporting restricted environments, the intentional bypass of transport security is a high-risk configuration.
Capability Assessment
Purpose & Capability
The name/description (local embedding setup on CPU-only machines behind an HTTP CONNECT proxy) matches the included instructions and script: probing connectivity, selecting a proxy, downloading a HuggingFace GGUF model via node-llama-cpp, verifying it, writing a .proxy record, and showing OpenClaw config/restart steps. Nothing requested by the skill (no external credentials, no unrelated service access) appears extraneous to that purpose.
Instruction Scope
Instructions and the script operate within the claimed scope: probing TLS and HTTP CONNECT proxies, writing a proxy record under ~/.openclaw/workspace/skills/openclaw-local-embedding/.proxy, and advising how to set openclaw.json and restart the gateway. Two points to be aware of: (1) the script and docs recommend setting NODE_TLS_REJECT_UNAUTHORIZED=0 for the download process (disables TLS certificate validation) — this is dangerous and should only be used if you trust the proxy performing MITM; (2) restart advice includes pkill -9 -f openclaw-gateway which force-kills processes and should be run carefully.
Install Mechanism
This is an instruction-only skill with a helper Node.js script included. There is no install spec that downloads arbitrary code or archives. Model download is performed via node-llama-cpp's resolver (HuggingFace model ID), and source-build fallback (llama.cpp) may compile locally if needed — this is coherent with the stated target environment.
Credentials
The skill does not require persistent credentials or declared environment variables. It does read existing HTTPS_PROXY environment variables if present and sets proxy-related env vars process-scoped during download. The recommended temporary setting NODE_TLS_REJECT_UNAUTHORIZED=0 weakens TLS validation during the process — acceptable in some corporate/proxy scenarios but risky otherwise. The script also contains a hard-coded Kuaishou cloud internal proxy (10.74.176.8:11080) which is reasonable for Kuaishou VMs but should be removed/ignored on non-Kuaishou hosts.
Persistence & Privilege
The skill does not request always: true, does not modify other skills, and only writes a small .proxy file under its own workspace. It does advise restarting the OpenClaw gateway (user action), but it does not assert permanent elevated privileges or force-inclusion.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install openclaw-local-embedding - After installation, invoke the skill by name or use
/openclaw-local-embedding - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.4
Finalize: confirm all steps verified, config validation with fallback, complete workflow
v1.0.3
Step 4: clarify config set version compat, add config get verification and validate check, add JSON fallback path on failure
v1.0.2
Review: remove redundant Step 1, add openclaw config set CLI method, broaden description, renumber steps
v1.0.1
Fix: remove manage.sh, use standard openclaw gateway restart, remove source-only --openclaw-dir placeholder
v1.0.0
Initial release: CPU-only local embedding setup for proxy-gated networks
Metadata
Frequently Asked Questions
What is Openclaw Local Embedding?
Initialize and configure OpenClaw local embedding mode on CPU-only machines. Handles network probe, proxy fallback, GGUF model download, cmake/llama.cpp comp... It is an AI Agent Skill for Claude Code / OpenClaw, with 153 downloads so far.
How do I install Openclaw Local Embedding?
Run "/install openclaw-local-embedding" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Openclaw Local Embedding free?
Yes, Openclaw Local Embedding is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Openclaw Local Embedding support?
Openclaw Local Embedding is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Openclaw Local Embedding?
It is built and maintained by LeeGoDamn (@leegodamn); the current version is v1.0.4.
More Skills