← Back to Skills Marketplace
Mistral Codestral
by
Twin Geeks
· GitHub ↗
· v1.0.2
· MIT-0
135
Downloads
0
Stars
2
Active Installs
3
Versions
Install in OpenClaw
/install mistral-codestral
Description
Mistral and Codestral — run Mistral Large, Mistral-Nemo, Codestral, and Mistral-Small locally. Mistral AI's open-source LLMs for code generation and reasonin...
Usage Guidance
This skill appears to do what it says: help you run Mistral/Codestral models locally via an ollama-herd router. Before you proceed: 1) Verify the upstream project (GitHub repo) and the PyPI package 'ollama-herd' authors/versions to ensure you trust the code you will pip install. 2) Run pip installs in a virtualenv or isolated system; inspect package contents if you can. 3) Be aware that running the 'herd' daemon creates a local HTTP service (default port 11435) — ensure it is bound to localhost and protected by your firewall to avoid exposing inference APIs. 4) Model downloads can be very large and will consume disk and network bandwidth; the SKILL.md claims downloads require explicit confirmation but validate that yourself. 5) The SKILL.md metadata references ~/.fleet-manager logs/db — these files may contain usage telemetry; review their contents and permissions. If you want more confidence, ask for the exact pip package metadata (author, homepage, SHA256 of the release) or request a repository snapshot to review the code before installing.
Capability Assessment
Purpose & Capability
The name/description claim local hosting of Mistral/Codestral models; the SKILL.md only asks for curl/wget and optionally python3/pip and shows examples talking to a localhost endpoint (http://localhost:11435). Those requirements are proportional. Minor mismatch: the registry metadata at top lists no required config paths, but the SKILL.md metadata includes configPaths (~/.fleet-manager/latency.db and ~/.fleet-manager/logs/herd.jsonl) — reasonable for a fleet router but inconsistent with the registry summary.
Instruction Scope
Runtime instructions stay within the stated purpose: install ollama-herd via pip, run herd/herd-node, and call a local HTTP API. Commands reference local endpoints and monitoring endpoints. The SKILL.md does not instruct reading unrelated system files or exfiltrating data to external endpoints. It does reference local config/log paths in metadata (see purpose_capability note).
Install Mechanism
There is no registry-level install spec, but the document instructs 'pip install ollama-herd' which pulls code from PyPI (moderate risk compared to no-install). This is expected for this purpose, but pip installs should be treated as untrusted code unless you verify the package source and integrity.
Credentials
The skill requests no environment credentials and only requires common network/CLI tools (curl/wget, optional python/pip), which is proportionate. The metadata's configPaths reference user-local fleet manager DB/logs; while plausible for a fleet router, you should be aware the fleet software may read/write those files and they could contain usage or telemetry data.
Persistence & Privilege
The skill is not forced always-on and is user-invocable; it allows autonomous invocation (the platform default). The main privilege is that running the recommended 'herd' daemon opens a local service (port 11435) to serve models — standard for local inference but worth securing and restricting to localhost/firewalls.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install mistral-codestral - After installation, invoke the skill by name or use
/mistral-codestral - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.2
Cross-platform support: macOS, Linux, and Windows. Updated OS metadata, descriptions, and hardware recommendations.
v1.0.1
- Updated documentation (SKILL.md) for improved clarity and emphasis on Mistral AI branding.
- Expanded multilingual and international description for broader audience reach.
- Reorganized setup, usage, and monitoring instructions with more examples focused on Mistral models.
- Added "Contribute" section and clarified guardrails for user safety.
- No changes to functionality; update is documentation only.
v1.0.0
Initial release: Run Mistral and Codestral models across your local fleet with optimized routing.
- Supports Mistral Large, Mistral-Nemo, Codestral, Mistral-Small, and Mistral 7B models.
- Enables code generation (Codestral) trained on 80+ programming languages.
- Routes model requests to the best device in your local network for efficiency.
- Simple OpenAI-compatible API, with endpoints for code, reasoning, image generation, speech-to-text, and embeddings.
- No automatic downloads; all model pulls and deletions require user confirmation.
- Includes web dashboard for monitoring, hardware fit guidance, and comprehensive documentation links.
Metadata
Frequently Asked Questions
What is Mistral Codestral?
Mistral and Codestral — run Mistral Large, Mistral-Nemo, Codestral, and Mistral-Small locally. Mistral AI's open-source LLMs for code generation and reasonin... It is an AI Agent Skill for Claude Code / OpenClaw, with 135 downloads so far.
How do I install Mistral Codestral?
Run "/install mistral-codestral" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Mistral Codestral free?
Yes, Mistral Codestral is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Mistral Codestral support?
Mistral Codestral is cross-platform and runs anywhere OpenClaw / Claude Code is available (darwin, linux, windows).
Who created Mistral Codestral?
It is built and maintained by Twin Geeks (@twinsgeeks); the current version is v1.0.2.
More Skills