Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add nebius/nebius-physical-ai --skill development-desktopgit clone --depth 1 https://github.com/nebius/nebius-physical-aiWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nebius/nebius-physical-ai/development-desktop)<a href="https://agentmods.dev/skills/nebius/nebius-physical-ai/development-desktop"><img src="https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/development-desktop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nebius/nebius-physical-ai/development-desktop"><img src="https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/development-desktop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00035 | $0.01224 |
| Opus 5.5 | $0.00014 | $0.00490 |
| Sonnet 5.5 | $0.00007 | $0.00245 |
| Haiku 4.5 | $0.00003 | $0.00122 |
Grade A, and why
development-desktop scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 28d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 83 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Development desktop
Use npa tools desktop for an operator's remote development environment.
Implementation belongs in npa/src/npa/tools/desktop, independently of Workbench.
- Inventory the exact existing VM and its SSH access. Preserve its checkout, running applications, credentials, and unrelated services. The current installer supports Ubuntu AMD64 and requires passwordless sudo.
- Run
npa tools desktop setup --ssh-host <alias> --dry-run --json, then the same command without--dry-run. This installs XFCE, TigerVNC, the pinned noVNC viewer, full VS Code, and the Codex extension. Reuse existing VNC and keyring passwords. The desktop stays on loopback unless public access is explicitly requested. - Open with
npa tools desktop open --ssh-host <alias>. Prefer one active viewer; old noVNC tabs can resize the shared desktop using CSS pixels and make Retina fonts look enormous. The viewer's More space control changes workspace scale while retaining lossless rendering and device-pixel density.display --dpi 192suits a Retina display; use 96 for standard density. For delayed clicks or typing, measure VM load and network latency, then useoptimize --ssh-host <alias>to disable compositing and GTK animations live. Preserve resolution and running applications; compare actual input-to-frame latency before claiming an improvement. Setup uses these same defaults. - For authorized public access, verify the VM's external IPv4 address and
ingress for TCP 80 and the selected HTTPS port. Then use
public-access --ssh-host <alias> --public-ip <address> --https-port 8443. Preserve existing port 443 services. This uses Certbot's short-lived IP certificates, automatic renewal, and a separate generated strong login. Passwords stay in private VM files, never URLs, Git, or PR evidence. - Prove trusted TLS, unauthenticated 401s for the page, WebSocket and credential endpoint, authenticated desktop rendering, and rejection of foreign origins. If ACME validation fails, investigate public routing before retrying; an address reachable from an operator network is not proof of Internet reachability.
status --jsonreports existing backup and snapshot records. Setup preserves recovery services but does not provision a backup repository or storage keys. Use encrypted backups with independent recovery keys, consistent online SQLite copies for Codex databases, and actual file/database restore checks.- For mobile control, run
chat-setup --ssh-host <alias> --connect-vscodeafter authenticated public access works. The mobile UI lives at/chat/on the same origin and uses the same desktop login. Show VS Code-origin history in Recent, Archived, and search; exclude CLI, execution, subagent, and unknown-origin chats without deleting them. Filter before pagination. New chat should offer existing recent project paths on that host, highlightnebius-physical-aicheckouts, and allow a custom path. Verify that selection controls the new chat's working directory and invalid paths remain editable. It shares a private Codex runtime with the VDI IDE; reload the IDE only when its work is idle. Keep sessions held by older independent Codex processes read-only until their owner releases them. Test real prompts and replies in both directions using the actual VS Code window and a phone-sized browser, plus reconnects, active-turn steering/stopping, approvals, and working tab/session indicators. Check that background completions replace the spinner with a persistent unread dot, including native status refreshes, and that only successfully viewing the finished reply clears it. Preserve dots across reloads and failed opens; do not mark existing idle history unread on first use. Verify shared model/reasoning controls against the live runtime catalog; restore test selections afterward. Preserve the existing account and model configuration. Never copy account tokens to the browser or restart a shared engine with active work during setup. - For existing Mac sessions, use
chat-setup --localand optionally--gateway-ssh-host <existing-managed-gateway>. Local mode follows the native VS Code owner, preserving active chats without changingchatgpt.cliExecutable. Require macOS, Node.js 22.13+, npm, and a signed-in Codex installation. Keep credentials in private runtime files. Preserve the existing HTTPS origin, credentials, desktop, and remote chat route. On a standalone mobile gateway, route old root bookmarks and Home Screen entry points to/chat/; preserve the original UI at/legacy/and its running services. Verify entry redirects, chat fragments, existing login cookies, and unsent-draft migration. Validate local/remote CLI exclusivity, same-thread messages, images, Plan/speed settings, draft retention, reconnects, repeated setup, and lost-response delivery without duplicate execution. Usestatus --local --jsonandopen --local --chatfor the saved setup. Recheck extension IPC compatibility after upgrades. Do not claim a backup exists from setup alone, or restart a mobile-owned active turn during updates.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 28d ago First seen · 83 lines · 35 tokens per session scan A 359594660f8f
development-desktop is a skill published in the GitHub repository nebius/nebius-physical-ai (50 stars, last pushed today), licensed Apache-2.0. It adds 35 tokens to every session and 1,224 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-24.
Other skills, from other repositories
ccal
Chinese Calendar with Lunar-Solar Conversion.
x-theme
Change terminal command line prompt theme. Globally manage the display style of all x-cmd interactive components. Supports multiple theme vendors: theme, starship, ohmyposh. Dependency: This is an x-cmd module. Install x-cmd first (see x-cmd skill for installation options). see x-cmd skill for installation.
loop
Schedule recurring tasks with notifications. Triggered by "/loop " to set up periodic monitoring.
atmos-custom-commands
Custom CLI commands: command definition in atmos.yaml, arguments, flags, shared native step types, when: conditions, output/UI steps, env vars, custom component types.
skill0
Root index of x-cmd skill0 sub-skills. Defines the OKR-style agent workflow (goal → rule-verified results → execute), skill discovery, and agent tooling preferences. Style: principle-first, concise, delegate specifics to authoritative external sources.
x-starship
Starship.rs cross-shell prompt with theme management. Minimal, blazing-fast, and infinitely customizable prompt. Auto-downloads starship binary if not available. Dependency: This is an x-cmd module. Install x-cmd first (see x-cmd skill for installation options). see x-cmd skill for installation.