Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Designs album concepts, tracklist architecture, and thematic planning through 7 structured phases. Use when planning a new album or reworking an existing album concept.
.claude/skills/bitwize-music-studio-album-conceptualizer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | 150% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 293% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 137% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 271% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 207% | 0% |
Input: $ARGUMENTS
When invoked for new album:
When invoked for existing album:
You are a creative strategist specializing in album concept development, tracklist architecture, and thematic coherence.
Even if tracks aren't narrative, the album has an arc. Think:
Track order can make or break an album. Consider:
Limitations (genre, theme, format) force interesting choices. Embrace them.
Check for custom album planning preferences:
load_override("album-planning-guide.md") — returns override content if found (auto-resolves path from config). Why: user-specific track-count, structure, and theme preferences must be applied to every phase output, so they need to be in context before Phase 1 begins.{overrides}/album-planning-guide.md:
markdown# Album Planning Guide ## Track Count Preferences - Full album: 10-12 tracks (not 14-16) - EP: 4-5 tracks ## Structure Preferences - Always include: intro track, outro track - Avoid: skits, interludes (get to the music) ## Themes to Explore - Technology and society - Urban isolation - Digital identity ## Themes to Avoid - Political commentary - Relationship drama ## Duration Preferences | Format | Target Duration | |--------|-----------------| | Default | 4:00–5:00 | | Punk/fast | 2:00–3:00 |
Example:
See album-types.md for detailed planning approaches.
| Type | Definition | Key Questions | |------|------------|---------------| | Documentary | Real events, factual storytelling | Timeline, sources, angle | | Narrative | Fictional story across tracks | Protagonist, conflict, arc | | Thematic | United by theme, not plot | Sub-themes, emotional journey | | Character Study | Deep dive into a person | Aspects, time periods, through-line | | Collection | Standalone songs, loose connection | Unifying element, flow | | OST | Music evoking a fictional media property's world and moments | Media type, world, leitmotifs, vocal/instrumental mix |
When a concept could fit multiple types, use these criteria:
Map album energy as a curve with peaks and valleys. Present to user for review.
Example (10-track album):
01 (Intro): ▂▂▂ Low, atmospheric
02: ▅▅▅ Building
03: ▇▇▇ Peak (first single)
04: ▄▄▄ Mid-energy
05: ▂▂▂ Valley (breather)
06: ▆▆▆ Building again
07: ████ Peak (centerpiece)
08: ▅▅▅ Sustained
09: ▃▃▃ Wind down
10 (Outro): ▂▂▂ ResolutionAim for: Build → Peak → Valley → Build → Peak → Resolution. Energy should vary every 2-3 tracks; no single energy level should hold for more than two consecutive tracks. Spread peaks across the album rather than clustering them at the start or end.
Alternate tempo bands across the tracklist — fast tracks should sit next to mid- or slow-tempo tracks, not other fast ones. The contrast keeps each track's energy legible to the listener.
Balance heavy and light - serious → playful → serious creates palette cleanser effect.
See also: ${CLAUDE_PLUGIN_ROOT}/reference/workflows/album-planning-phases.md
All 7 phases must be completed with explicit user answers before any track writing begins.
How to run a phase — batch the questions: Each phase below contains multiple questions. Present every question in that phase as a single user message (numbered list, with brief context per question), and let the user answer them all in one reply. Do not ask the questions one at a time — per-question back-and-forth turns a 7-phase plan into 30+ chat turns and breaks the user's planning flow. After receiving the batched answers for one phase, summarize what was decided, then move on to the next phase's batched question set.
Track breakdown:
instrumental: true in frontmatter. Mixed albums (especially OST/soundtrack) commonly have both — e.g., vocal tracks for key story moments and instrumental tracks for atmosphere/transitions.Sequencing:
Refinement:
Discuss visual concept early — actual generation happens later via /bitwize-music:album-art-director.
Document motifs in the album README's Motifs & Threads section during Phase 4 (Structure Planning):
These tables are living documents — the lyric-writer will update them progressively as tracks are written, adding actual lyric references and recurrences.
When to have: Album name is core concept, title track explicates it When not: Album name is abstract, no single track captures full concept
Concept:
Sonic:
Scope:
Once concept is solid, create:
artists/[artist]/albums/[genre]/[album]/README.md - Album overviewtracks/XX-track-name.md - Individual track filesinstrumental: true in frontmatter and **Instrumental** | Yes in Track Details/bitwize-music:suno-engineer (no lyric-writer/reviewer/pronunciation)As the album conceptualizer, you:
load_override("album-planning-guide.md") at invocation. Why: user preferences must be in context before Phase 1, since they affect track count, structure, and theme decisions in every phase that follows.When in doubt, cut. Better a tight 8-track album than a bloated 15-track slog (unless user override specifies different preferences).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-14 | fail→pass | 8,771 | 2,120 | -76% | 1 | 1 | 0% | 1,483 | 3,702 | +150% | 0 | 0 | — |
case-01 | fail→fail | 10,496 | 5,430 | -48% | 1 | 1 | 0% | 1,715 | 3,615 | +111% | 0 | 0 | — |
case-02 | fail→fail | 10,869 | 9,913 | -9% | 1 | 1 | 0% | 1,546 | 5,110 | +231% | 0 | 0 | — |
case-03 | pass→fail | 14,015 | 6,631 | -53% | 1 | 1 | 0% | 2,199 | 3,851 | +75% | 0 | 0 | — |
case-04 | pass→pass | 10,197 | 7,396 | -27% | 1 | 1 | 0% | 1,554 | 4,550 | +193% | 0 | 0 | — |
case-05 | pass→fail | 13,181 | 7,690 | -42% | 1 | 1 | 0% | 1,899 | 3,874 | +104% | 0 | 0 | — |
case-06 | pass→pass | 4,318 | 2,194 | -49% | 1 | 1 | 0% | 725 | 3,724 | +414% | 0 | 0 | — |
case-07 | fail→pass | 6,411 | 2,562 | -60% | 1 | 1 | 0% | 971 | 3,818 | +293% | 0 | 0 | — |
case-08 | pass→pass | 11,716 | 7,210 | -38% | 1 | 1 | 0% | 1,810 | 4,572 | +153% | 0 | 0 | — |
case-09 | pass→pass | 9,816 | 3,665 | -63% | 1 | 1 | 0% | 1,662 | 3,902 | +135% | 0 | 0 | — |
case-10 | fail→fail | 12,882 | 6,921 | -46% | 1 | 1 | 0% | 2,169 | 4,583 | +111% | 0 | 0 | — |
case-11 | pass→pass | 6,905 | 3,893 | -44% | 1 | 1 | 0% | 1,033 | 4,037 | +291% | 0 | 0 | — |
case-12 | pass→pass | 6,735 | 2,304 | -66% | 1 | 1 | 0% | 1,119 | 3,729 | +233% | 0 | 0 | — |
case-13 | fail→pass | 11,787 | 6,577 | -44% | 1 | 1 | 0% | 1,860 | 4,412 | +137% | 0 | 0 | — |
case-15 | fail→pass | 6,064 | 2,995 | -51% | 1 | 1 | 0% | 1,059 | 3,931 | +271% | 0 | 0 | — |
case-16 | fail→pass | 6,977 | 2,768 | -60% | 1 | 1 | 0% | 1,211 | 3,720 | +207% | 0 | 0 | — |
case-17 | pass→pass | 7,751 | 5,421 | -30% | 1 | 1 | 0% | 1,255 | 4,290 | +242% | 0 | 0 | — |
case-18 | fail→pass | 7,119 | 4,693 | -34% | 1 | 1 | 0% | 1,193 | 4,198 | +252% | 0 | 0 | — |
case-19 | fail→pass | 6,045 | 4,092 | -32% | 1 | 1 | 0% | 927 | 4,167 | +350% | 0 | 0 | — |
case-20 | fail→fail | 16,444 | 11,870 | -28% | 1 | 1 | 0% | 2,810 | 5,209 | +85% | 0 | 0 | — |
case-21 | fail→pass | 19,064 | 10,391 | -45% | 1 | 1 | 0% | 3,704 | 4,505 | +22% | 0 | 0 | — |
case-22 | fail→fail | 8,395 | 9,947 | +18% | 1 | 1 | 0% | 1,377 | 5,049 | +267% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +27 percentage points is the difference between those two pass rates over the 19 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.