New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Wally Device E2E

by RunanywhereAI in RunanywhereAI/wally

Run wally's LLM e2e on Apple Neural Engine (NeuRT) and Snapdragon Hexagon NPU (QHexRT) devices.

MITAuto-check passedTesting & QA

Install Wally Device E2E

skills CLI
$ npx skills add RunanywhereAI/wally --skill wally-device-e2e -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install RunanywhereAI/wally wally-device-e2e --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/RunanywhereAI/wally.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/wally-device-e2e .claude/skills/wally-device-e2e && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
wally-device-e2e
GitHub stars
1.6k
Token cost
~845 tokens
SKILL.md length
351 words
Files
1
Skills in repo
5
Repo updated
First seen
Licence
MIT

At a glance

Run wally's LLM e2e on Apple Neural Engine (NeuRT) and Snapdragon Hexagon NPU (QHexRT) devices.

  • Adding overlay backends
  • SKILL.md covers Run, Local model ids and Overlay gotchas
  • Calls bash
  • Proving LLM inference on device

What it does

Wally Device E2E is an agent skill from RunanywhereAI/wally. Run wally's LLM e2e on Apple Neural Engine (NeuRT) and Snapdragon Hexagon NPU (QHexRT) devices. Use when adding overlay backends, proving LLM inference on device, or when a PC only has one backend's bundles on disk. Non-LLM modalities (STT/TTS/VLM/embed/image/VAD/rerank/segment/diarize) are deferred while the LLM-only cut is in effect.

Its SKILL.md is about 850 tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Testing & QA, covering End-to-end testing, Text to speech and voice and LLM inference and serving. It works with Qwen. The repository describes itself as: Get up and running with GLM-5.3-flash, DeepSeek, Gemma and other open source frontier models. The licence is MIT.

When your agent uses it

  • Adding overlay backends
  • Proving LLM inference on device
  • A PC only has one backends bundles on disk

Example prompts

  • “/wally-device-e2e”

What it can do on your machine

Read from SKILL.md and the folder at commit 39b923e. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • bash

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Wally Device E2E loads about 845 tokens when it runs. Until then it costs about 89 tokens; SKILL.md has 351 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~89
When it runs · the whole SKILL.md, loaded when a task matches
~845

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from RunanywhereAI/wally at commit 39b923e, republished under its MIT licence (© RunanywhereAI). 351 words, ~845 tokens.

Download SKILL.mdSave it as .claude/skills/wally-device-e2e/SKILL.md (or your agent's skills folder).
name
wally-device-e2e
description
Run wally's LLM e2e on Apple Neural Engine (NeuRT) and Snapdragon Hexagon NPU (QHexRT) devices. Use when adding overlay backends, proving LLM inference on device, or when a PC only has one backend's bundles on disk. Non-LLM modalities (STT/TTS/VLM/embed/image/VAD/rerank/segment/diarize) are deferred while the LLM-only cut is in effect.

Wally device modality e2e

Do not write per-engine tests. The harness is scripts/test/e2e-modalities.sh, called from scripts/test/e2e.sh, keyed by primitive (llm, stt, tts, vlm, embed, image, vad, rerank, segment, diarize). This build's LLM-only cut (src/app.rs) registers only the llm command — every other primitive's wally subcommand is commented out, so pointing the harness at one now fails with "no such command", not a skip. Run llm only until that cut is lifted. wally picks the engine from catalog framework, local path, or plugin priority. --engine is an override (WALLY_E2E_ENGINE), never a required test input.

Run

bash
# Public CI (modelless): skip every modality
bash scripts/test/e2e.sh /path/to/wally

# Device: discover whatever is already on disk, then run llm
export RUNANYWHERE_HOME=/path/to/home          # already-pulled OSS models
export WALLY_E2E_MODEL_ROOTS=/path/to/hnpu:/path/to/coreml
bash scripts/test/e2e-modalities.sh /path/to/wally

# Or pin the model explicitly (path or catalog id)
WALLY_E2E_LLM=/path/to/lfm2_5_230m_HNPU \
  bash scripts/test/e2e-modalities.sh /path/to/wally

WALLY_E2E_AUTO=1 also sets defaults for stt/tts/vlm/embed/vad/ rerank/segment/image (whisper-tiny, piper, minilm, …); on this LLM-only cut every one of those now fails with "no such command" instead of skipping, since their wally subcommand does not exist. Only the llm default (smollm2 / mlx-qwen3) actually runs — treat any other AUTO failure as the disabled command, not your change. Never enable AUTO in public CI.

Local model ids

The Windows ARM64 box often only has LFM *_HNPU trees under Downloads\hnpu, copied for LLM smoke. Catalog ids:

QHexRT id (local *_HNPU)NeuRT id (local Core ML tree)
lfm2_5_230mlfm2_5_230m_ane

wally models pull of a Hugging Face repo page is HTML. Pass the expanded directory to -m. Download v81/* only on Hexagon v81.

Skip with a clear "no bundle" when the tree is missing. Fail only when a model was selected and the command failed.

Show full SKILL.md (120 more words)Show less

Overlay gotchas

  • Public bottles never list neurt / qhexrt. Rebuild product wally against an overlay kit (WALLY_SDK_KIT pointing at that prefix).
  • QHexRT: QAIRT 2.48 on Snapdragon X2 Elite / Hexagon v81. ADSP_LIBRARY_PATH must be the fully expanded ...\lib\hexagon-v81\unsigned path. Nested %QNN_SDK_ROOT% in cmd /c set does not expand. Copy QnnHtp*.dll next to wally.exe. FastRPC ~90s then user-driver fallback is normal. Use a .bat, not nested cmd /c.

Non-LLM overlay coverage (NeuRT image generation, llama.cpp VLM, segment, STT) is deferred, not deleted: those primitives run through this same harness once src/app.rs's LLM-only cut is uncommented, but until then their wally subcommands do not exist, so this skill does not instruct running them.

See wally-e2e for bottle/backends assertions and Apple MLX host link flags.

© RunanywhereAI, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .agents/skills/wally-device-e2e of RunanywhereAI/wally.

Open the folder on GitHubat commit 39b923e

Compare with similar skills

Wally Device E2E next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Wally Device E2E compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Wally Device E2E this skillRunanywhereAI/wally1.6k—~845Automated safety check: PassMIT
Quark Torch LLM Ptq Evalamd/Quark181—~2.6kAutomated safety check: PassMIT
Qwen Code E2E TestingQwenLM/qwen-code28k—~2.1kAutomated safety check: PassApache-2.0
tmux Real User TestingQwenLM/qwen-code28k—~2.3kAutomated safety check: PassApache-2.0
Generate Profilesgl-project/sglang37k2 repos~1.1kAutomated safety check: PassApache-2.0
Run Agmente E2Erebornix/Agmente545—~414Automated safety check: PassMIT

Similar skills

  • Quark Torch LLM Ptq Eval

    amd/Quark

    L3 recipe that runs a Torch LLM PTQ end-to-end for AMD Quark — for PyTorch / HuggingFace transformers models (safetensors input): quantize → validate → evaluate.

    181 GitHub stars~2.6k tokensUpdated 10 days ago
    AI & LLM EngineeringAuto-check passed

  • Qwen Code E2E Testing

    QwenLM/qwen-code

    Guides end-to-end testing of the Qwen Code CLI in headless mode with real model calls, MCP test servers and inspection of raw API traffic.

    28k GitHub stars~2.1k tokensUpdated today
    Testing & QAAuto-check passed

  • tmux Real User Testing

    QwenLM/qwen-code

    Drives Qwen Code in a real tmux session the way a user would and saves a readable step-by-step transcript of each screen for maintainers to review.

    28k GitHub stars~2.3k tokensUpdated today
    Testing & QAAuto-check passed

  • Generate Profile

    sgl-project/sglang

    Generate an e2e profiling trace of an SGLang server run. An agent skill from sgl-project/sglang.

    37k GitHub starsUsed in 2 repos~1.1k tokens
    Testing & QAAuto-check passed

  • Run Agmente E2E

    rebornix/Agmente

    Run Agmente iOS end-to-end tests against a local ACP agent (Gemini, Claude, Qwen, or Vibe), validate core RPC flow, and perform mandatory cleanup.

    545 GitHub stars~414 tokensUpdated 4 mo ago
    Testing & QAAuto-check passed

  • Demo Video

    hmislk/hmis

    A skill your agent uses when asked to make a demo, training, how-to or tutorial video with sound or voice-over showing an HMIS function or configuration (e.g.

    236 GitHub stars~3.9k tokensUpdated today
    Testing & QAAuto-check passed

More from RunanywhereAI/wally

  • Wally Architecture

    RunanywhereAI/wally

    Where Wally logic belongs — command layering, proto as SOT, kit vs CLI ownership, Apple MLX host vs wally-cxx.

    1.6k GitHub stars~992 tokensUpdated today
    Auto-check passed

  • Wally Release

    RunanywhereAI/wally

    Cut an Wally product release (independent of SDK version) — version bump, release:patch label, merge, auto-tag, bottles.

    1.6k GitHub stars~1.3k tokensUpdated today
    Auto-check passed

  • Wally E2E

    RunanywhereAI/wally

    Verify a built wally binary against a pinned C++ desktop kit on macOS and Windows.

    1.6k GitHub stars~3.8k tokensUpdated today
    Auto-check passed

  • Wally Kit Pin

    RunanywhereAI/wally

    Bump cmake/sdk-pin.cmake to a new published SDK C++ desktop kit (version + SHA-256 + IDL lock).

    1.6k GitHub stars~1k tokensUpdated today
    Auto-check passed

Works with

Categories

Questions about Wally Device E2E

What does Wally Device E2E do?

Run wally's LLM e2e on Apple Neural Engine (NeuRT) and Snapdragon Hexagon NPU (QHexRT) devices. Wally Device E2E is an agent skill from RunanywhereAI/wally. Run wally's LLM e2e on Apple Neural Engine (NeuRT) and Snapdragon Hexagon NPU (QHexRT) devices.

When should I use Wally Device E2E?

Wally Device E2E fits situations like: adding overlay backends; proving LLM inference on device; A PC only has one backends bundles on disk.

How do I install Wally Device E2E in Claude Code?

Run `npx skills add RunanywhereAI/wally --skill wally-device-e2e -a claude-code`. Or copy the skill folder (.agents/skills/wally-device-e2e in RunanywhereAI/wally) into .claude/skills/wally-device-e2e in your project. Claude Code loads it when a task matches its description.

How do I install Wally Device E2E in Codex?

Run `npx skills add RunanywhereAI/wally --skill wally-device-e2e -a codex`. Or copy the skill folder (.agents/skills/wally-device-e2e in RunanywhereAI/wally) into .agents/skills/wally-device-e2e in your project. Codex loads it when a task matches its description.

Can I use Wally Device E2E in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add RunanywhereAI/wally --skill wally-device-e2e -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/wally-device-e2e, .gemini/skills/wally-device-e2e, .github/skills/wally-device-e2e and .opencode/skills/wally-device-e2e in your project.

What does Wally Device E2E need to run?

Going by SKILL.md and its folder, Wally Device E2E needs the command-line tools its instructions call (bash).

Does Wally Device E2E access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Wally Device E2E safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Wally Device E2E use?

Wally Device E2E is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Wally Device E2E use?

About 845 tokens (SKILL.md is roughly 3.4k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Wally Device E2E?

Skills that share tags, products or a category with Wally Device E2E: Quark Torch LLM Ptq Eval (amd/Quark, 181 stars), Qwen Code E2E Testing (QwenLM/qwen-code, 28k stars), tmux Real User Testing (QwenLM/qwen-code, 28k stars) and Generate Profile (sgl-project/sglang, 37k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Wally Device E2E?

RunanywhereAI (a GitHub organization) maintains it in RunanywhereAI/wally, which has 1,564 GitHub stars. The repository holds 5 skills in this directory. The repository was last updated on October 8, 2026.

Source: RunanywhereAI/wally on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.