Music track skill for AI agents

music-track is a skill from the Summer skill library: a guide your AI agent reads when the task matches, so it builds the good version on the first try. Generate a music track: game loops of 30s or more with a loop point, or linear cinematic cues.


Status	Stable
Domains	Audio, Asset pipeline
Version	`1.0.2`
Engine version	`>=4.6`
Recommended set	Yes, installed by `--recommended`

When your agent uses it

Generating a looped or cinematic music track — main theme, boss music, title screen, exploration loop, combat track
A game has no music yet or needs a new track for a level, a cutscene, or the credits

Install

summer setup installs every skill for your agent. To add only this one:

npx -y summer-engine@latest skills install music-track --agent claude-code

Swap claude-code for your agent, such as codex or cursor. See Skills for every option.

The skill

Overview

ElevenLabs Music renders a track from a prompt. Summer clamps the requested length to 3–600 seconds before sending it, so anything outside that range is silently pulled to the nearest bound. Quality depends on prompt specificity: tempo (BPM), key, instrumentation, mood, structural cue. “Epic orchestral” returns generic stock music. “85 BPM, D minor, solo cello and detuned upright piano, sparse, melancholic, no drums, 4/4” returns a usable score for a melancholic exploration scene. This skill produces one track aligned with the audio bible’s music style and dynamic music plan, then wires it as an AudioStreamPlayer on the Music bus with a marked loop point if it’s a loop, or as a linear cinematic stream if it’s a one-shot.

When to use

Looped track for a level, biome, menu, or game state (calm / tension / combat).
Cinematic linear track for an intro, ending, boss reveal, or cutscene.
Replacing a placeholder loop with on-bible music.

When NOT to use

Short stings / event punctuation (<5s) → use sound-effect with class Stinger.
Adaptive layered stems with state-driven crossfades → adaptive-music.
Ambient bed with no melody → ambient-bed.

Steps

1. Read the audio bible

Read .summer/audio-bible.md

The bible names the primary style, tempo range, key center, time signature, and the dynamic music model. This skill writes prompts that respect those numbers. If no bible exists:

No .summer/audio-bible.md. Music without an audio bible drifts into stock-library territory. Run /audio-direction first, or I can ship a one-off with explicit defaults — your call.

2. Identify the track’s role

Ask if not obvious:

Which slot in the dynamic music plan? Calm exploration, tension, combat, boss, victory, menu, ending, or one-off cutscene?

Each slot has different rules:

Slot	Tempo	Energy	Instrumentation density	Loop or linear
Calm exploration	bible low end (e.g. 70–90)	Low	Sparse	Loop, ≥45s
Tension	mid (e.g. 90–110)	Medium, no drums or sparse pulse	Adds one instrument	Loop, ≥45s
Combat	high (e.g. 120–150)	High, full drums	Full	Loop, ≥45s
Boss	high or low-and-menacing	Distinct from combat	Distinct, named instrument	Loop, ≥60s
Victory sting	n/a	Resolution	n/a	Linear, 4–8s (use SFX skill instead)
Menu / title	bible mid	Atmospheric	Anchor instrument	Loop, ≥30s
Cutscene	varies	Linear arc	Varies	Linear, 60–180s
Ending	varies	Linear arc with resolution	Varies	Linear, 60–300s

3. Build the prompt — tempo + key + instrumentation + mood + structural cue

Prompts that work:

85 BPM, D minor, 4/4, solo cello and detuned upright piano, sparse, melancholic
exploration loop, no drums, gentle reverb, even dynamics, 60s

128 BPM, D dorian, 4/4, full orchestral combat, low brass ostinato, taiko drums
on 1 and 3, strings tremolo, urgent, no melodic resolution, 60s loop

90 BPM, A minor, 6/8, lo-fi exploration, vintage Rhodes electric piano,
tape hiss, warm bass, soft brushed snare, no vocals, 45s loop

70 BPM, F minor, 4/4, ambient menu loop, pad layer, distant choir, single
piano motif every 8 bars, no drums, 30s loop with seamless return

110 BPM, C minor harmonic, 4/4, dark synthwave boss intro, gated reverb snare,
arpeggiated bass synth, lead saw on melody, builds to cymbal swell at 0:50,
linear 60s, ends on sustained pad

Prompts that DON’T work (and why):

Bad prompt	Failure mode
`epic orchestral combat music`	No tempo, no key, no instrumentation. Generic stock.
`Hollow Knight style track`	The model can’t reference IPs and won’t try to clone style words.
`hopeful but sad menu music with strings and stuff`	”Strings and stuff” is vague; model picks the wrong section.
`something for the boss fight`	No tempo, no key, no mood vector. Random.
`loop` (just the word)	No structure. Returns 30s of music that doesn’t loop.

4. Tempo, key, instrumentation specifics that matter

Tempo (BPM) — name a single number, not a range. Music model interprets ranges as mid-point with drift. 85 BPM is better than 80–90 BPM. Key — name mode (D minor, D dorian, D natural minor, C minor harmonic). The bible’s modulation rules go here: combat tracks shift to dorian, calm tracks to natural minor. Instrumentation — name 2–4 specific instruments, not sections. solo cello not strings. upright piano not keys. Rhodes electric piano not electric piano. The model’s vocabulary is closer to a session musician’s than a producer’s. Mood — concrete adjectives anchored to motion: melancholic and slow, urgent and driving, triumphant and rising. Avoid genre adjectives like epic, chill, hype. Structural cue — for linear tracks, name the arc: builds at 0:30 to a hit at 0:50, decays. For loops, name the cell: 4-bar phrase repeats, instrument swap every 8 bars.

5. Loop vs cinematic decision

Choose loop when	Choose linear when
The slot plays for an unbounded time (level, menu, exploration)	The slot has a known length (cutscene, intro, ending)
Player attention is on gameplay	Player attention is on the music
You want to layer adaptive stems on top later	The arc is the point

For loops: generate ≥30s, prefer 45–60s. Shorter loops are recognized as repetition within 90s of play and become annoying. The bible’s “no state stays >90s without variation” rule means even a 60s loop should contain internal variation (instrument swap on alternating cycles). For linear: generate ≥60s for cutscenes, up to 300s for endings. Ask for one rendered piece rather than splicing; the model handles long arcs.

6. Confirm and call

Show the prompt:

Prompt: 85 BPM, D minor, 4/4, solo cello and detuned upright piano, sparse, melancholic exploration loop, no drums, gentle reverb, 60s loop Duration 60s. This uses credits. Generate?

The music capability reads the description from prompt, not text — the mirror image of sound_effects. Without prompt the call is refused (prompt_required).

assets.generate_audio({
  capability: "music",
  prompt: "85 BPM, D minor, 4/4, solo cello and detuned upright piano, sparse, melancholic exploration loop, no drums, gentle reverb, 60s loop",
  durationSeconds: 60
})
// Then: summer_import_from_url(url: "<asset.fileUrl>", path: "res://audio/music/calm_exploration.mp3")

durationSeconds is clamped to 3–600, and a longer track costs more: dryRun: true returns the estimate in credits first. Music comes back as MP3 (audio/mpeg), so a .mp3 target path is correct.

7. Mark the loop point in Summer Engine

ElevenLabs Music does not return a loop-clean cut. You must mark the loop point in Summer Engine’s Import dock so playback returns to the loop point on completion, not to 0:00. The MP3 importer exposes exactly five options — loop, loop_offset, bpm, beat_count, bar_beats. There is no loop_mode on MP3 or OGG; loop_mode belongs to the WAV importer, where it is spelled edit/loop_mode. In audio/music/calm_exploration.mp3.import:

[params]
loop = true
loop_offset = 0.0

If the start has a fade-in but the body is loopable, set loop_offset = 4.0 (4 seconds of intro skipped on loop, plays full track first time). bpm + beat_count + bar_beats are what you set if you want beat-synced transitions later — see adaptive-music. For seamless looping, the cleaner option is to ask the model to generate a cell (one bar / one phrase) and concatenate; or generate at 60s and trust the model’s structural cue. Crossfade looping in code is the practical fallback (step 9).

8. Wire it as `AudioStreamPlayer` on the `Music` bus

One script adds the player with its settings; scenePath is the exact scene to change:

summer_run_script(scenePath="res://main.tscn", source="""
func run(ctx):
    ctx.add_node("AudioStreamPlayer", "MusicCalm", null, {
        "stream": load("res://audio/music/calm_exploration.mp3"),
        "bus": "Music",
        "volume_db": -8.0,
        "autoplay": true,
    })
    ctx.save_scene()
""")

Music is AudioStreamPlayer (not 2D / 3D) — non-positional. The bible’s Music -8 dB mix sits the music behind SFX where it belongs.

9. Crossfade-loop pattern (no marked loop point)

If the loop click bothers the user, crossfade two players. Set loop = false in the import file first, so each player stops at the end.

# scripts/audio/MusicLoopCrossfade.gd
extends Node
@export var stream: AudioStream
@export var fade := 1.0
@export var volume_db := -8.0
var _players: Array[AudioStreamPlayer] = []
var _current := 0

func _ready() -> void:
	for i in 2:
		var p := AudioStreamPlayer.new()
		p.stream = stream
		p.bus = "Music"
		p.volume_db = volume_db
		add_child(p)
		_players.append(p)
	_players[0].play()

func _process(_delta: float) -> void:
	var a := _players[_current]
	if not a.playing or a.get_playback_position() < stream.get_length() - fade:
		return
	_current = 1 - _current
	var b := _players[_current]
	b.volume_db = -80.0
	b.play()
	var t := create_tween().set_parallel()
	t.tween_property(a, "volume_db", -80.0, fade)
	t.tween_property(b, "volume_db", volume_db, fade)
	t.chain().tween_callback(a.stop)

The next player starts fade seconds before the current one ends and the two cross over, so the seam is never heard.

Reference card — prompts by slot

Calm explore (folk):    78 BPM, D minor, 6/8, fingerpicked acoustic guitar,
                        upright bass on 1, soft hand percussion, no melody on
                        first 8 bars, 60s loop

Tension (orchestral):   95 BPM, D dorian, 4/4, low strings sustained, single
                        timpani roll every 8 bars, sparse french horn motif,
                        no drums, 60s loop

Combat (orchestral):    140 BPM, D dorian, 4/4, taiko drums on 1 and 3,
                        ostinato low brass, strings tremolo, urgent and driving,
                        no melodic resolution, 60s loop

Boss (synthwave):       110 BPM, C minor harmonic, 4/4, gated snare, arpeggio
                        bass synth, lead saw on melody, builds at 0:50,
                        linear 60s

Menu (lo-fi):           70 BPM, F minor, 4/4, vintage Rhodes, tape hiss,
                        soft brushed snare, single piano motif every 8 bars,
                        45s loop

Cutscene (cinematic):   varies tempo, D minor, 4/4, solo piano opens,
                        strings enter at 0:20, full orchestra at 0:50,
                        decay to solo piano at 1:30, linear 120s

Anti-patterns

Genre adjectives only. epic, chill, hype — meaningless.
Naming a game or composer. Model won’t replicate; use ingredients.
Loop shorter than 30s. Recognized as repetition fast.
Music on Master bus at 0 dB. Buries SFX. Use Music bus at -8 dB.
Generating before reading the bible. Style drift.
Skipping the loop-mark step. Music returns to 0:00 with a click on every loop.

Edge cases

Clip is 50s but you wanted 60s. Either regenerate or extend by overlap-crossfading the last 5s onto the start in an audio editor.
Loop click on every cycle. Use the crossfade pattern in step 9.
Clip is too long. Trim the tail in the import dock with loop_offset, or in an audio editor.
Track sounds nothing like the references. Re-read the bible; the prompt may be missing the anchor instrument. Add it explicitly.
Track has unexpected vocals. Add no vocals, instrumental only to the prompt.

Without Summer Engine

In a session without Summer Engine tools (the assets profile; the summer-assets skill has the whole loop), do steps 1-6 and skip steps 7-9 (loop point, AudioStreamPlayer, crossfade script). Save the file where the person asked with out (the summer-assets skill says how), and tell them the path and format. For a loop track, tell them it is made to loop.

Fallback (no MCP)

Print the prompt and instruct the user to run via the Summer dashboard, then summer_import_from_url the rendered MP3.

Handoff

Track calm_exploration.mp3 wired to Game/MusicCalm on the Music bus, autoplay on. Next:

Mark the loop point in the import dock if the bible calls for looped playback.

Generate the tension and combat counterparts with /music-track again.

To wire state-driven crossfades between calm / tension / combat, run /adaptive-music.

From a shell

No Summer MCP tools in this session? Run each one as a command; when summer --version is missing or below 4.0, use npx -y summer-engine@latest in place of summer. Fields are flags in kebab-case (--scene-path res://main.tscn) or one --input '<json>', and summer ops describe <op> prints the exact input. The summer-cli skill has the rest.

In this skill	Command
`summer_assets_generate_audio`	`summer assets generate-audio`
`summer_assets_search`	`summer assets search`
`summer_add_node`	`summer engine add-node`
`summer_import_from_url`	`summer engine import-from-url`
`summer_inspect_node`	`summer engine inspect-node`
`summer_run_script`	`summer engine run-script`
`summer_set_prop`	`summer engine set-prop`

adaptive-music: Wire music stems to game state: combat, explore, boss and tension crossfades on a shared bus.
audio-direction: Define the game’s sound (music style, SFX vocabulary, dynamic music) as .summer/audio-bible.md.

Source: library/skills/music-track/SKILL.md, rendered from summer-engine@4.3.0 on npm.

​When your agent uses it

​Install

​The skill

​Overview

​When to use

​When NOT to use

​Steps

​1. Read the audio bible

​2. Identify the track’s role

​3. Build the prompt — tempo + key + instrumentation + mood + structural cue

​4. Tempo, key, instrumentation specifics that matter

​5. Loop vs cinematic decision

​6. Confirm and call

​7. Mark the loop point in Summer Engine

​8. Wire it as AudioStreamPlayer on the Music bus

​9. Crossfade-loop pattern (no marked loop point)

​Reference card — prompts by slot

​Anti-patterns

​Edge cases

​Without Summer Engine

​Fallback (no MCP)

​Handoff

​See also

​From a shell

​Related skills