opensite-ai/opensite-skills / linux-distro-selector
linux-distro-selector
VerifiedChoose the right Linux distribution for a developer or AI/ML workstation and confirm the hardware will work before installing. Use when someone asks which Linux distro to pick; is comparing Ubuntu, Fedora, Arch, NixOS, Pop!_OS, Mint, or atomic/immutable distros; is switching from Windows or macOS and unsure what to install; needs a distro matched to a use case (maximum stability, bleeding-edge GPU/AI work, reproducibility, laptop battery life, or a server); or wants to check whether a GPU, Wi-Fi chipset, or laptop is Linux-compatible. Advisory only: it produces a recommendation, not install commands.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.8 KB |
| agents/openai.yaml | config | 402 B |
| examples/nvidia-laptop-stability.md | markdown | 4.0 KB |
| references/activation.md | markdown | 2.9 KB |
| references/atomic-vs-traditional.md | markdown | 4.8 KB |
| references/distro-matrix.md | markdown | 11.8 KB |
| references/hardware-compat-triage.md | markdown | 6.6 KB |
| templates/hardware-intake-worksheet.md | markdown | 2.7 KB |
examples/nvidia-laptop-stability.md
View on GitHub# Worked Example — NVIDIA Laptop, Wants Stability + Recent CUDA
A worked walk-through of the advisory flow. Versions are **as of 2026-07 — verify**
at runtime before presenting them to the user.
---
## User request
> "I just got a Lenovo Legion laptop with an RTX 4070 (mobile). I'm an AI/ML
> engineer. I want the machine to be **stable and low-maintenance**, but I also
> need **reasonably recent CUDA and PyTorch**. What Linux distro should I install?"
## Step 1 — Elicit primary optimization + constraints
Filled from [../templates/hardware-intake-worksheet.md](../templates/hardware-intake-worksheet.md):
- **Primary optimization:** never-breaks / low-maintenance — but with a hard secondary
requirement for recent CUDA/PyTorch. Tie-break axis leans toward *stability*, with
freshness delivered through tools, not the OS.
- **GPU:** NVIDIA RTX 4070 mobile (Ada / Ampere-successor, i.e. **Turing+** → open kernel
modules are the default). Compute needed (CUDA + PyTorch).
- **Form factor:** laptop, almost certainly **NVIDIA Optimus** hybrid (Intel iGPU + NVIDIA dGPU).
- **Wi-Fi:** Intel AX-series (smooth, in-kernel) — no wired-install flag.
- **Secure Boot:** user wants to keep it on, does not want to manage keys.
- **Container comfort:** yes, comfortable doing dev in containers.
## Step 2 — Triage the hardware
Against [../references/hardware-compat-triage.md](../references/hardware-compat-triage.md):
- **NVIDIA laptop → Optimus/PRIME handling required.** Favor a distro that smooths this over.
- **Secure Boot on + no key management → out-of-box shim distro** (rules out Arch/CachyOS
unless they accept `sbctl`).
- **Turing+ GPU** → open kernel modules; Wayland + NVIDIA is daily-usable on the 5xx/59x driver
series. No X11-only pro tools mentioned, so Wayland is fine.
- Wi-Fi and battery: `power-profiles-daemon`/`tuned` available on the leading candidates.
## Step 3 — Walk the decision tree
From SKILL.md: "never break" + "NVIDIA laptop" + "comfortable with containers" points to two
branches — the **atomic** path and the **stable-LTS-with-fresh-tools** path. Both satisfy
"stable host, recent CUDA via tools/containers."
## Step 4 — Recommendation
**Primary pick: Bluefin (Universal Blue, NVIDIA image)** — atomic Fedora-based dev workstation.
- **Rationale:** immutable host that can't be broken by a bad driver/update (roll back at boot in
seconds), NVIDIA + Optimus handled by the dedicated ublue nvidia image, Secure Boot works out of
the box, and all bleeding-edge CUDA/PyTorch work happens in **distrobox** (e.g. a CUDA container)
so the host stays pristine. Best "stable + fresh dev" combination for this user.
- **Biggest trade-off:** you do GPU/ML dev inside containers, not on the bare host — a small
mental-model shift, and occasional friction with tools that expect a mutable FHS root.
**Alternative: Ubuntu 26.04 LTS** — traditional mutable system.
- **When to prefer it:** if the user wants a conventional, widely-documented system that matches
cloud/CI images, or dislikes container-first dev. Ubuntu 26.04 now ships **NVIDIA CUDA natively
in its repos** with signed Secure Boot support and `ubuntu-drivers`, so recent CUDA is one command
away; pull newer PyTorch via `uv`/`pixi` or containers. Trade-off: a mutable host is easier to
break than an atomic one, and you handle Optimus via `prime-select` yourself.
Also viable: **Pop!_OS 24.04** (System76-tuned NVIDIA + tiling) or **Bazzite** (NVIDIA + Mesa
pre-installed) if the user wants the absolute least NVIDIA setup effort.
## Step 5 — Verify + hand off
- **Verify** the current Bluefin nvidia image tag, Ubuntu 26.04 support window, and the current
NVIDIA driver series for the RTX 4070 with a quick web search before finalizing.
- **Next step:** hand the confirmed hardware profile (NVIDIA RTX 4070 mobile, Optimus, Intel Wi-Fi,
Secure Boot on) to the `linux-install-planner` skill to generate the install runbook, then use
`linux-dev-workstation` to configure the NVIDIA/PRIME drivers, and `linux-ai-dev-stack` for the
local + cloud AI toolchain.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |