grok-build with the leash off — one binary, any OpenAI-compatible endpoint, zero phone-home. No x.ai account.
# Add to your Claude Code skills
git clone https://github.com/xiaojumaoaihuaban/grok-unleashedGuides for using ai agents skills like grok-unleashed.
grok-unleashed is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by xiaojumaoaihuaban. grok-build with the leash off — one binary, any OpenAI-compatible endpoint, zero phone-home. No x.ai account. It has 50 GitHub stars.
grok-unleashed's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/xiaojumaoaihuaban/grok-unleashed" and add it to your Claude Code skills directory (see the Installation section above).
grok-unleashed is primarily written in Shell. It is open-source under xiaojumaoaihuaban on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh grok-unleashed against similar tools.
No comments yet. Be the first to share your thoughts!
Unlocks once the catalog security scan passes (runs nightly).
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
The grok-build CLI with the leash off.
One static binary. Any OpenAI-compatible endpoint. Zero x.ai dependency, zero telemetry, zero upload queues, zero permission nags. It runs where you run, talks to whoever you point it at, and ships your prompts to exactly one place: the API you configured. Nowhere else.
The official build wants an x.ai OAuth session before it will say hello. It phones home — telemetry, source-artifact uploads, workspace snapshots. It interrupts you with permission dialogs mid-flow. And outside the first-party API it treats every other endpoint like a second-class citizen, tripping over its own streaming bugs until relays 502.
I stripped all of it. 11 files. -597 lines, +185 back. The diff is mostly deletion, which is the honest kind of security work: the fastest code to audit is code that no longer exists.
| Removed | How you know |
|---|---|
| Source-artifact upload worker | never spawns in production paths — locked by unit tests that ship with the build |
| Workspace snapshot collection | upload queue constructor discards the queue itself |
| Telemetry / OTLP export | no exporter endpoints, no trace upload env is honored |
| Permission prompts / sandbox / managed policy / folder trust | no dialogs, no policy fetch, no trust gating |
| First-party credential routing | kill switch at the credential seam — nothing ever routes to api.x.ai, even by accident |
Third-party endpoints aren't just allowed — they work properly:
base_url + model + key and you're on
DeepSeek, Zhipu, Kimi, ollama, or any relay. Context windows up to 1M.Grab the binary from releases, then:
1. Install (Linux or WSL, pick one):
# easy way — download install.sh from the release too, then:
sudo bash install.sh
# manual way:
sudo install -m 0755 grok-unleashed-*-linux-x86_64 /usr/local/bin/grok
2. Give it a key — any OpenAI-compatible API:
mkdir -p ~/.grok
nano ~/.grok/config.toml
Paste this, replace base_url / model / api_key with yours:
[models]
default = "mine"
[model.mine]
model = "deepseek-v4-pro"
base_url = "https://api.deepseek.com/v1"
api_key = "sk-..."
api_backend = "chat_completions"
context_window = 128000
Works with: DeepSeek, Zhipu, Kimi, ollama (local), any relay that
speaks /chat/completions. More examples in config.example.toml.
3. Run it:
grok
That's it. First run asks about workspace mode once — pick "always current directory" and it never asks again.
Windows users: install inside WSL (wsl --install -d Ubuntu-24.04),
then call it from PowerShell via wsl -- grok or drop a one-line
grok.cmd (@wsl.exe -d Ubuntu-24.04 -- grok %*) somewhere in PATH.
Every claim above is checkable in one line each:
# where does it actually connect?
strace -f -e trace=connect grok -p "hi" --output-format plain
You'll see your endpoint (and your local proxy, if any). You will not see api.x.ai, grok.com, assets.grok.com, or any hub — not even a DNS lookup that resolves.
# is anything queued for upload?
find ~/.grok -path '*upload_queue*' -type f | wc -l
0
Always zero. Restart recovery purges spilled queue files on boot — the failure mode where data waits on disk for a chance to leave doesn't exist here.
# is the binary what it claims?
sha256sum grok-unleashed-*-linux-x86_64
49ac228ef43844def90f773c787fc75cd03d6d91ccfe5500fa9e3662d73b2048
grok update. It fetches the official binary and
everything above comes back.This is an unofficial patched build. The excellent original CLI is
xai-org/grok-build
(Apache-2.0, base commit 8adf901, version 0.2.101). Patches here
are published under the same license. Not affiliated with xAI.