# Mlx Fleet

> Preview and confirm a one-shot per-role router configuration.

- Skill: `cavi-ai/mlx-fleet-2` (Agent Skill, multi-file: 45 files)
- Install (CLI): `npx skillmds@latest add cavi-ai/mlx-fleet-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/cavi-ai/mlx-fleet-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: cavi-ai (https://skillmd.com/u/cavi-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/cavi-ai/mlx-fleet-2

---


Use `$mlx-agent:mlx-fleet` to invoke this installed Codex skill explicitly. Codex does not support custom `/mlx-*` slash commands.

# MLX Fleet

Resolve `<skill-dir>` as the absolute directory containing this SKILL.md. Never resolve the bundled executable from the shell working directory.

canonical capability ID: mlx-agent.fleet

Render the one-shot per-role router configuration without mutation:

`python3 <skill-dir>/scripts/mlx-agent fleet render --path <router.yaml> --assign <role=repo> --json`

Then request the exact transaction diff and preview hash without confirmation:

`python3 <skill-dir>/scripts/mlx-agent fleet apply --path <router.yaml> --assign <role=repo> --json`

Show that returned diff and preview hash. Do not write configuration files directly. Only after the user explicitly confirms that exact preview, run:

`python3 <skill-dir>/scripts/mlx-agent fleet apply --path <router.yaml> --assign <role=repo> --confirm --preview-hash <preview-hash> --json`

Never download model weights without an explicit confirmation. Report the transaction receipt returned by the CLI.