Turn astory/plotinto a song with an arc of development: produce Style tags + structured Lyrics, and guide the user to copy and paste them into suno.com to generate. Suno has no official API; this skill only produces text, and the final generation relies on the user manually operating on the official website.
Core engine (must be loaded first, execute according to its skeleton)
The engine of this skill is the "Suno AI Music Architect" Narrative Music Architect method.
The first thing after triggering: load NARRATIVE_ENGINE.md, and execute according to it — use its skeleton and process, do not restate, do not generalize.
Its crux is Structural Mapping (structure mapping protocol): map the story's arc of development onto musical sections (Intro→Verse→Pre-Chorus→Chorus→Bridge→Outro). Narrative mapping is themain axis, do not dilute it.
When executing, overlay the factual patch:SUNO_REFERENCE.md(the methodology took shape in v3.5, and the patch brings Style/Lyrics up to the current v5.5 state,without touching the engine skeleton). For version/factual information, the patch prevails and may be updated.
Language and generation
- Language: by default lyrics are in Chinese / Style tags in English (Suno recognizes English tags most accurately). If the user clearly uses English or wants an English song, use all English.
- Rhythm: output the full set at once according to the engine's Output Format (analysis report + Style + Lyrics). If not satisfied, iterate.
- Lyrics self-check: after producing the Lyrics, you must perform a rhyme/singability self-check according to
SUNO_REFERENCE.mdSection 6 before delivering — lock down the rhymes in the chorus, and read each section aloud to smooth out the final words. - Wrap-up: after output, attach the suno.com operation flow (see
SUNO_REFERENCE.mdSection 4) — turn on Custom, paste the first section into the Style box, and paste the second section into the Lyrics box.
Input handling
- Has a story/plot → go directly through the engine; narrative mapping is the main line.
- Insufficient information(only says "write a song" without giving material) → use the engine's Initialization questions to ask back: What is the story outline? What core emotion do you want to convey?
- Pure instrumental / BGM(says "make a BGM" and does not want lyrics) → still follow the engine to do Style; in the Lyrics section keep only the structure tags +
(instrumental), and add at the end of Styleinstrumental。
Extension mode (not the main axis, enabled only when the input is genuinely not a story)
When the input is a single emotion, a theme, or a piece of copy rather than a narrative,only thenrelax the engine's "story parsing" — arrange sections according to the emotional arc, while the rest (style anchoring, Output Format, the section logic of structural mapping) still follows the engine. This is a fallback, not the default; if it can be leaned toward narrative, lean toward narrative.
Reference resources
NARRATIVE_ENGINE.md— Core engine, load it first on every trigger, and execute according to its skeleton.SUNO_REFERENCE.md— v3.5→v5.5 factual patch + tag lexicon + official website flow, overlaid during execution.
Cross-platform
Plain text output, no scripts/dependencies/credentials, consistent across Mac/Windows/Linux. The user's only external action is to open suno.com in a browser and paste manually.