suno-skills

name	suno-skills
description	Create, analyze, rewrite, optimize, and troubleshoot Suno music generations. Use for Suno Style prompts, Simple or Custom mode briefs, lyrics and section tags, instrumental arrangements, Exclude lists, Voices or Personas, Covers, audio references, Custom Models, prompt variants, controlled A/B tests, and diagnosing generations that missed their intended sound.

Suno Skills

Turn an idea or existing production into a controlled Suno generation package. Optimize for audible decisions, coherent priorities, original expression, and useful iteration rather than maximal prompt length.

Route the request

1. Identify the job:
   • Create: produce a Suno-ready package from an idea.
   • Rewrite: improve an existing Style, Exclude, or lyrics block.
   • Diagnose: explain why a generation missed and propose the smallest next test.
   • Translate: convert a reference track or production vocabulary into audible Suno language.
   • Iterate: design controlled variants and a listening scorecard.
2. Read only the references required for that job:
   • Prompt construction: references/prompt-design.md
   • Lyrics, prosody, and section tags: references/lyrics-structure.md
   • Genre, fusion, instruments, vocals, and production vocabulary: references/genre-palette.md
   • Instrument layering, vocal design, mood, production, and Exclude: references/arrangement-controls.md
   • Named subgenre fingerprints and drift repairs: references/genre-deep-dives.md
   • Failures, repair order, and A/B testing: references/troubleshoot-evaluate.md
   • Current Suno product features and source policy: references/suno-product.md

Gather the creative target

Infer safe omissions. Ask at most three short questions only when the missing answer would materially change the song. Establish, when relevant:

• purpose and target duration;
• primary musical lane and optional secondary color;
• energy, motion, tempo feel, and emotional arc;
• prominent sound sources and their roles;
• lead voice or instrumental focus;
• production space and era;
• lyric language, perspective, subject, hook, and boundaries;
• elements to exclude;
• current Suno model or feature when the workflow depends on it.

Never pretend BPM, key, exact duration, singer identity, or section timing is guaranteed. Present them as steering cues unless the active product explicitly exposes deterministic controls.

Build the generation package

Use this priority order:

1. Primary lane.
2. Motion and energy.
3. Two to four defining sound sources with roles.
4. Voice or lead behavior.
5. Production space and dynamic arc.
6. Optional era, region, meter, BPM, or key cue.

Prefer concrete, audible phrases: fingerpicked nylon guitar, dry rim-click pocket, close-mic restrained alto, narrow verse opening into a wide chorus. Remove praise words such as amazing, professional, or beautiful unless translated into sound.

Use one dominant genre and no more than one supporting lane by default. Resolve contradictions instead of averaging them. Keep positive instructions in Style and unwanted sonic elements in Exclude when that control is available. Enforce lyrical topics, imagery, safety, and cultural boundaries while writing the lyrics; do not dump them into Exclude.

Treat a reference-list term as a search lead, not a finished prompt. When the user names a narrow style, read references/genre-deep-dives.md, identify its audible fingerprint, and convert that fingerprint into motion, source roles, lead behavior, and production arc.

When the request depends on layer balance, duet/choir behavior, section-local vocal changes, emotional atmosphere, mix space, or stubborn unwanted elements, read references/arrangement-controls.md. Keep global sonic identity in Style and put only local section behavior in lyrics tags.

For Custom mode, return the applicable fields in this order:

Title:

Style:

Exclude:

Lyrics / Structure:

Advanced controls:

Why this should work:

Next A/B test:

Omit empty fields. Keep Why this should work to one or two sentences. If the user asks only for a copy-ready prompt, return only the fields they need.

Handle lyrics

Read references/lyrics-structure.md. Keep section tags on their own lines. Treat bracketed cues as probabilistic arrangement hints, not commands. Preserve repeated chorus wording when hook consistency matters. Match syllable density to pace and leave space for phrasing.

Do not imitate or continue copyrighted lyrics. If a user supplies an artist or song reference, extract high-level musical traits and produce original wording. Do not include artist names in the final Suno prompt unless the user explicitly needs them for cataloging; prefer a trait translation.

Diagnose and iterate

Read references/troubleshoot-evaluate.md. Diagnose the largest mismatch first:

1. lane;
2. motion;
3. palette;
4. lead or vocal behavior;
5. structure and lyric density;
6. production and finish.

Change one major variable per test. Keep lyrics, model, seed-like controls, and other settings stable where possible. Generate equal sample counts, score only the target trait, and save exact inputs.

Offer two kinds of variants only when useful:

• Controlled variant: one deliberate change for diagnosis. Express it as an exact add/remove/replace patch. Do not restate unchanged traits inside the replacement phrase.
• Creative variant: a clearly labeled alternative interpretation.

Never claim a prompt phrase is universally proven from one generation.

Verify current product behavior

Suno features and limits change. Before giving instructions about current models, plan entitlements, limits, rights, UI labels, Voices, Personas, Custom Models, uploads, Covers, Studio, or commercial use, browse current official Suno documentation. Use unofficial sources only for practical heuristics and identify them as such. See references/suno-product.md.

Use the helper script

Run scripts/suno_prompt.py when deterministic assembly or linting is useful:

python scripts/suno_prompt.py build brief.json
python scripts/suno_prompt.py lint --style "..." --exclude "..."

The builder accepts JSON fields documented by python scripts/suno_prompt.py schema. Treat its output as a strong draft, then apply musical judgment.

Quality gate

Before delivery, confirm:

• Every Style phrase maps to something audible.
• The primary lane is obvious in the first phrase.
• Genre, motion, instruments, voice, and production do not conflict.
• Instrument nouns have roles or textures where useful.
• Rhythm, harmony, lead, and texture layers have an intelligible priority.
• Exclude does not negate required Style elements.
• Lyrics fit the requested perspective, language, pace, and section shape.
• The prompt does not promise exact control Suno may not provide.
• The package includes a small, testable next move rather than an indiscriminate rewrite.