Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/kobiton/automatenpx agentmods add skills/kobiton/automate/debug-device-over-vusbWrote 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/kobiton/automate/debug-device-over-vusb)<a href="https://agentmods.dev/skills/kobiton/automate/debug-device-over-vusb"><img src="https://agentmods.dev/badge/skills/kobiton/automate/debug-device-over-vusb/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/kobiton/automate/debug-device-over-vusb"><img src="https://agentmods.dev/badge/skills/kobiton/automate/debug-device-over-vusb.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 | $0.00176 | $0.07030 |
| Opus 5.5 | $0.00070 | $0.02812 |
| Sonnet 5.5 | $0.00035 | $0.01406 |
| Haiku 4.5 | $0.00018 | $0.00703 |
Grade A, and why
debug-device-over-vusb 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 3d 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 — 241 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Debug on a Real Device over virtualUSB
Overview
Attach a real Kobiton device to the user's own machine over virtualUSB and debug the user's app on it the way they would with a phone on their desk: it shows up in adb devices (Android) or Xcode's device list (iOS), so the user's normal install / reproduce / read-the-logs loop works, driven from plain language. The skill owns the whole lifecycle - client install, device discovery, connect, verification, the debug loop, and an unconditional disconnect - and reports every failure as a plain-language meaning plus the next step.
Use it when the user wants to debug or reproduce a problem locally against a real device: "debug my app on a real Pixel", "make the device show up in adb", "get logcat from a real device while I reproduce this", "see the device in Xcode", "plug a Kobiton device into my machine", or any mention of "virtualUSB" / "vusb". For exploratory testing through Kobiton's own session CLI, route to run-interactive-session instead; for an agent-driven flow that should become a test case, drive-automation-session.
Prerequisites
Runs on macOS (any architecture) and Windows (x64 under Git Bash) and needs a persistent local filesystem - it runs a locally cached client binary, reads ~/.kobiton/.credentials, and hands device access to local adb / xcrun. Linux hosts are redirected by the preflight (outcome=redirected (limitation)): tell the user to use a macOS or Windows machine, or run-interactive-session for device access from Linux. See the Skill compatibility matrix in CLAUDE.md.
Before the flow can succeed:
- virtualUSB client - installed by this skill's own preflight (Step 1), never by the plugin's SessionStart hook or
/automate:setup, so only users of this skill download it. The plugin pins one client build inskills/debug-device-over-vusb/VUSB_VERSION; the preflight fetches that build fromhttps://public.kobiton.download/virtualusb/<version>/, verifies its published sha256, and on macOS caches the wholevirtualUSB.appbundle under~/.kobiton/vusb/<version>/. On Windows it caches the verified.msiand prints the install steps for the user (UAC, thenvusb setup-adbin an administrator terminal). The wrapper~/.kobiton/bin/vusb(symlink on macOS, exec-shim on Windows) resolves the binary by absolute path -~/.kobiton/binis not onPATHand must not be added. - Credentials file -
~/.kobiton/.credentialswithKOBITON_USER,KOBITON_API_KEY,KOBITON_PORTALin the active profile ($KOBITON_PROFILE, defaultdefault), written by/automate:setup.~/.kobiton/bin/vusb loginreads it and signs the client in with the API key; the key never appears in the transcript. - Kobiton MCP connection -
listDevices({virtualUsb: true, …})is how vUSB-ready devices are found. MCP tool names below are bare (listDevices); the host resolves its registered prefix. - Local tooling -
adbonPATHfor Android; Xcode command line tools (xcrun devicectl) on macOS for iOS. On Windows, adb must be installed before the one-timevusb setup-adb, and an administrator terminal that runs as another account needs adb's folder added to its ownPATH(the preflight's hand-off names it). - Account and devices - an organization with the virtualUSB add-on and at least one private device whose host machine has virtualUSB enabled with reachable routing (
virtual_usb_ready: true). Public cloud devices are never offered. - Administrator approval once - on macOS the first
connect(and every client version bump) installs the client's background daemon via a GUI administrator dialog; over SSH or without a GUI session this cannot complete. On Windows the.msineeds UAC.
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 3d ago Changed · +1 lines 0b5e067f3bd2
- 18d ago First seen · 240 lines · 176 tokens per session scan A 2547952c9949
debug-device-over-vusb is a skill published in the GitHub repository kobiton/automate (14 stars, last pushed 4d ago), licensed MIT. It adds 176 tokens to every session and 7,030 once invoked, about $0.0007 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 2026-10-02.
Other skills, from other repositories
xcodebuildmcp-cli
Official skill for the XcodeBuildMCP CLI. Use when doing iOS/macOS/watchOS/tvOS/visionOS work (build, test, run, debug, log, UI automation).
ios-simulator
Verify and debug native, React Native, Expo, or Flutter apps on an iOS Simulator with agent-device. Use when an agent needs to launch an app, inspect its live UI, tap, type, scroll, validate a code change, collect failure evidence, or reproduce a workflow on an iPhone or iPad Simulator.
spm-build-analysis
Analyze Swift Package Manager dependencies, package plugins, module variants, and CI-oriented build overhead that slow Xcode builds. Use when a developer suspects packages, plugins, or dependency graph shape are hurting clean or incremental build performance, mentions SPM slowness, package resolution time, build…
xcode-project-analyzer
Audit Xcode project configuration, build settings, scheme behavior, and script phases to find build-time improvements with explicit approval gates. Use when a developer wants project-level build analysis, slow incremental builds, guidance on target dependencies, build settings review, run script phase analysis…
phone-harness
Use when an agent needs to control a real phone — iPhone via Mac iPhone Mirroring or Android via adb (open apps, tap, type, swipe, OCR the screen). Android path runs on this Linux workstation (adb), targets phonefarm devices. iOS path needs macOS + iPhone Mirroring. Use for phonefarm device debugging, OTP/2FA flows…
rn-agent-observer
Runtime observability for React Native/Expo apps on Android. Use when debugging RN UI, performance (FPS, JS blocking), network, re-renders, or runtime errors on a running app — understand the current screen and UI findings, capture screenshots, UI trees, ref snapshots, per-request network, CDP console/heap/CPU…