list_presets
Play demoTry asking
- What sound presets do I have in Noice?
- list my noice presets
- which ambient mixes are saved
- show sleep sound presets
ghost skill run ghost/noice list_presets Noice ambient sounds (no account): list presets and sounds, play a preset for a few seconds then stop, stop playback.
Task check: passed without the skill, passed with it.
One recorded run each way on list_presets, claude-haiku-5-5, same phone and same prompt. Watch it in the Before / after tab.
14.4× fewer tokens, 1.3× slower with the skill. Task check: passed without the skill, passed with it.
| Run | Time | Tokens | Tool calls | Result |
|---|---|---|---|---|
| No skill: agent drives by hand | 15.1 s | 296,293 | 10 | passed |
With skill: find_skill, run_skill (list_presets) | 19.4 s | 20,578 | 3 | passed |
Prompt: List the available sound presets in Noice
WITH skill was 4.31s slower despite fewer calls. Replay trace: open_presets 10.262s plus list_presets 5.662s (run_skill total 15.951s); discovery and final response add the remaining elapsed time. Both sides list the same four presets. Source inspection shows the list collector rechecks pages and performs four reset swipes after collecting rows; it trades thorough collection for elapsed time.
One recorded pair on samsung-demo, claude-haiku-5-5; tokens count every input, cache and output token once per assistant message. Not a signed run: it adds no run evidence.
Demo shows: play_preset (preset=Beach, seconds=3)
Ghost in the Droid seed skill for Noice (
com.github.ashutoshgngwr.noice), an ambient sound generator. No account. Install it from F-Droid.
| Action | What it does | Params |
|---|---|---|
list_presets | Saved presets with author | none |
list_sounds | Every library sound with its category | none |
play_preset | Play a preset briefly, then stop | preset, seconds (3, max 30) |
stop_playback | Stop whatever Noice plays | none |
curl -LO https://f-droid.org/repo/com.github.ashutoshgngwr.noice_72.apk && adb install com.github.ashutoshgngwr.noice_72.apk
pip install "ghost-in-the-droid>=1.6.0" # 1.6.0 is not on PyPI yet: until it ships, git clone the repo and `pip install -e .`
ghost skill install ghost/noice
ghost skill run ghost/noice list_presets --device emulator-5554
ghost skill run ghost/noice play_preset --param preset=Thunderstorm --param seconds=3 --device emulator-5554
From any MCP agent:
run_skill(name="ghost/noice", action="play_preset", params={"preset": "Beach", "seconds": 5})
# {"played": "Beach by @eMPee584", "seconds": 5}
Android 14 emulator (google_apis x86_64, 1080x2400, 420 dpi), Noice 2.5.7, en-US, fresh install and reused state. Signed self-runs live in runs/.
Never opens the Account tab or signs in. Playback is short and always stopped. Needs network on first play of each sound. License: MIT (the app itself is GPL-3.0).
| App version | Android | Device | Skill version | Signed self-runs (pass/total) | Dates |
|---|---|---|---|---|---|
| 2.5.7 | 14 | Google/sdk_gphone64_x86_64 (emulator, 1080x2400, 420 dpi, en-US) | 1.0.0 | 8/8 | 2026-10-05 |
Physical-device launch attempts are also recorded on Samsung SM-A156B (Android 16, 1080×2340, en-US). Results remain action-specific; an attempt does not imply success. See the demos and current signed runs for evidence.
Untested so far: other OEM skins, other screen sizes and other locales.
None in the signed self-runs.
play_preset fails with "did not start playing".id/premium_status) need a subscription; the bundled presets play without one.SOFT.md is the narrative an agent reads when the exact steps no longer match your app version and it has to adapt the skill.
What each action means, which UI landmarks it relies on, and how success is recognised, so an agent can adapt it to another ambient-sound app.
Package: com.github.ashutoshgngwr.noice (Noice 2.5.7 from f-droid.org, sha256 verified against the repo index)
Verified: 2026-10-05 on a headless emulator pool, Android 14 google_apis x86_64, 1080x2400 @ 420 dpi, en-US, no account (the Account tab is never opened).
Skill class: HARD. No login, no checkpoints.
Welcome) with id/skip (SKIP) and id/next.id/library, id/presets, id/sleep_timer, id/alarms, id/account.LIFE, PLACES, RAW NOISE, THINGS, TRAVEL, WEATHER); each sound row has id/title, id/info, id/download, id/play, id/volume.id/play_button and id/title (Beach by @eMPee584) and id/menu_button.id/playback_state (Playing, then Stopping during the fade-out), id/play_toggle, id/stop.dumpsys media_session, package com.github.ashutoshgngwr.noice): state=PLAYING(3) or STOPPED(1), description=<preset name>.list_presets()Cold-start, Presets tab, return {presets: [{preset, name, author}], count}.
list_sounds()Cold-start, Library tab, walk all categories: {sounds: [{sound, category}], count} (31 sounds in 2.5.7).
play_preset(preset, seconds)Find the preset row by full title or by the name before by @, tap its play button, wait until the session plays that preset, wait seconds, tap the controller's Stop. Success: the session reaches STOPPED (the app fades out for a few seconds first).
stop_playback()If the session is playing, send KEYCODE_MEDIA_STOP. Success: not playing.
play_preset fails with "did not start playing".id/premium_status) need a subscription; the bundled presets play without one.Try a sample prompt or run an action by name. Replace any parameters in braces with your own values.
list_presetsTry asking
ghost skill run ghost/noice list_presets list_soundsTry asking
ghost skill run ghost/noice list_sounds play_presetTry asking
ghost skill run ghost/noice play_preset stop_playbackTry asking
ghost skill run ghost/noice stop_playback Each release is an immutable git tag. The tree hash pins the exact skill directory the installer verifies.
| Version | Released | App versions | Tree | Size | Provenance |
|---|---|---|---|---|---|
1.0.0latest | never | >=2.5.7 | pending | 19.3 KB | untagged |