| name | suno |
| description | Write song lyrics formatted for Suno AI V5.5 music generation and submit them via the `suno` CLI, which drives Suno through the user's own logged-in browser session on suno.com. Use when the user wants to create a song, write lyrics, compose music with AI, submit to Suno, list Suno personas, check Suno credits, or asks for help with Suno. Triggers on requests like "write me a song", "create lyrics about...", "help me with Suno", "compose a track", "song about...", "suno credits", "suno personas". |
| allowed-tools | bash |
Suno Songwriting Assistant
Write lyrics formatted for Suno AI's Custom Mode with proper metatags, structure, and style prompts.
End-to-End Workflow
- Clarify — ask about genre, mood, theme, and vocal preference if not provided
- Draft lyrics — use metatags, parameterized section headers, correct meter and rhyme
- Write style prompt — structured colon format or producer run-on sentence
- Run prosody audit — line-by-line syllable counts, rhyme scheme map, singability check
- Suggest sliders — Weirdness, Style Influence, Audio Influence values
- Confirm before submitting — show the user the final lyrics, style prompt, sliders, and target persona, then wait for explicit user approval.
suno generate and UI submission consume paid Suno credits, so never submit on your own initiative
- Submit — only after explicit approval, via
suno CLI (preferred) or UI automation fallback
- Poll and iterate —
suno poll <clip_id> --wait, then refine weak sections with Song Editor
How Suno Actually Works
Critical insight: Suno maps text into a probabilistic style-mesh rather than reading prompts literally.
Genre Defaults and Escaping Pop Gravity
Nearly every genre gravitates toward "pop" unless actively countered (rock→pop, funk→pop, emo→pop). Genre clusters also blur: asking for "boom bap" still pulls trap elements due to training co-occurrence.
Escape strategies:
- Explicit exclusions: "no trap", "no pop"
- Force unusual combinations: "emo industrial", "math rock gospel"
- Strategic contrast: emphasize elements that naturally oppose unwanted defaults
Output Format
Always output these components:
1. Lyrics (in markdown code block)
[Intro]
(instrumental description if needed)
[Verse 1]
Lyrics here with proper meter
Each line should have consistent syllables
Rhymes enhance but don't force them
[Chorus]
The hook that repeats exactly
Every time it appears
(background vocal echo)
[Verse 2]
Continue the narrative
Match Verse 1's syllable count
[Bridge]
Tonal or thematic shift
[Chorus]
The hook that repeats exactly
Every time it appears
(background vocal echo)
[Outro]
Closing lines or instrumental fade
2. Style Prompt
Use the colon-and-quotes format for maximum clarity—this is a parsing hint, not cosmetic:
genre: "indie folk rock, 2020s bedroom pop"
vocal: "soft female alto, intimate whisper-to-belt, gentle vibrato, slight nasal quality"
instrumentation: "fingerpicked acoustic guitar, warm upright bass, sparse piano, light ambient pads"
production: "lo-fi intimacy, tape warmth, close-miked vocals, narrow stereo image, natural room reverb"
mood: "melancholic, nostalgic, late-night introspection"
For simpler prompts, use producer-style run-on sentences:
A fusion of 80s dark synthwave and modern cyberpunk aesthetics with extremely high-fidelity recording and professional mastering and analog warmth and controlled high-end and phase-coherent low end.
Up to ~1,000 characters in V5.5. Aim for 4–7 descriptors (~60–200 chars) for cleanest results; longer prompts give more control but risk competing descriptors.
3. Creative Sliders (when relevant)
Suggest values for Suno's Custom Mode sliders:
Weirdness: 50% (0=conventional/radio-safe, 100=experimental/chaotic)
Style Influence: 70% (0=loose interpretation, 100=strict adherence to style prompt)
Audio Influence: 50% (only with reference audio; 0=inspiration only, 100=close mirror)
Style Prompt Formatting Rules
These rules apply to every style prompt you write.
Use Periods and "and/with", Not Commas
Suno sees commas as opportunities to skip what follows. Use "and" and "with" to create run-on sentences, and use periods to end conceptual units.
Wrong:
acoustic guitar, male vocals, emotional, reverb
Right:
acoustic guitar with male vocals and emotional delivery and reverb-heavy production.
Avoid Lyric Bleed
Suno performs soft classification between conditioning text and performable text. Anything singable might get sung.
Triggers to avoid in style prompts:
- Short poetic lines
- Brackets that look like stage directions
- ALL CAPS slogans
- Quoted phrases that could be lyrics
- Empty lyrics box
- Rhythmic prose
Keep style prompts metadata-like and dense: technical descriptions do not scan as lyrics.
Never Use Artist Names
Suno rejects prompts containing artist names (e.g., "Vangelis", "Beatles", "Billie Eilish"). Instead, describe the sonic characteristics:
| Instead of | Use |
|---|
| "Vangelis style" | "80s synth soundtrack, lush pads, cinematic, orchestral electronics" |
| "Beatles style" | "60s British Invasion, jangly guitars, vocal harmonies, Merseybeat" |
| "Billie Eilish style" | "dark pop, whispered vocals, minimal bass-heavy production" |
| "Johnny Cash style" | "sparse country, baritone vocals, acoustic guitar, train beat" |
Formatting Rules
Structure Tags (in square brackets)
[Intro] [Verse] [Pre-Chorus]
[Chorus] [Post-Chorus] [Bridge]
[Outro] [Hook] [Break]
[Fade Out] [Instrumental]
Parameterized Metatags (V5+)
Combine section tags with per-section production cues using colon or pipe syntax:
[Verse 1: whispered vocals, acoustic guitar only]
Lyrics here...
[Chorus: full band, anthemic, layered harmonies]
Hook lyrics...
[Bridge: piano only, vulnerable vocals, half-tempo]
Contrasting lyrics...
Or use the pipe format for stacked styling that overrides the global prompt:
[Chorus | anthemic chorus | stacked harmonies | modern pop polish]
Hook lyrics here...
[guitar solo | 80s glam metal lead | heavy distortion | whammy bar bends]
[Verse 2 | raspy lead vocal | emotional build-up | lo-fi warmth]
This gives per-section control over instrumentation, vocal style, and energy without putting production notes in the style prompt.
Vocal/Performance Tags
[Male Vocal] [Female Vocal] [Duet]
[Choir] [Spoken Word] [Whisper]
[Harmonies] [Raspy Voice] [Operatic]
Sound Effects (use sparingly)
[Cheering] [Clapping] [Whispers]
[Screams] [Sighs] [Chuckles]
MAX Mode (Acoustic/Folk/Orchestral)
For acoustic, country, folk, singer-songwriter, and orchestral work, MAX Mode dramatically improves quality:
[Is_MAX_MODE: MAX](MAX)
[QUALITY: MAX](MAX)
[REALISM: MAX](MAX)
[REAL_INSTRUMENTS: MAX](MAX)
[START_ON: TRUE]
[START_ON: "write out the first few words of lyrics here"]
genre: "outlaw country, 70s singer-songwriter"
instruments: "single dreadnought acoustic, baritone male, vocal fry, blue notes, melismatic runs"
style tags: "tape saturation, close-mic presence, small room acoustics, handheld mic grit, dry & raw"
Note: MAX Mode has minimal effect on electronic/trap/hip-hop/synthwave—use structural prompting for those.
Realism and Synthesis Descriptors
For detailed descriptor vocabulary, see references/descriptors.md. Key principles:
Acoustic/organic genres: use recording-engineer language (e.g., "close mic presence", "natural timing drift", "tape saturation", "short room reverb") instead of abstract vibes. Describe the recording environment and performance artifacts.
Electronic genres: shift away from "realism" toward synthesis and modulation language (e.g., "FM synthesis bass", "LFO-driven movement", "resonant bandpass motion", "mono-stable low end"). Describe motion, harmonic shape, and stereo control rather than size or weight.
Writing Guidelines
Meter and Prosody (Suno-Specific)
Key principle for Suno: stressed syllables must fall on musical beats for Suno to render words correctly. Unstressed syllables between beats can vary — this creates natural flow. When the natural word stress conflicts with the beat, Suno will mispronounce or mangle the line.
- Keep syllable counts consistent within sections (e.g., all verse lines at 8 syllables)
- Mark stresses when drafting to verify alignment:
da-DUM da-DUM da-DUM da-DUM
- Avoid padding with filler words ("it", "just", "so") to hit counts — restructure instead
- Genre affects expected meter: NDW/industrial favors 8–12 syllables, punchy and staccato; French chanson favors octosyllabic or alexandrine lines
Rhyme
- Chorus must rhyme — this creates the hook. Use perfect rhymes ("night/light") for maximum impact.
- Verses can use slant rhyme ("home/alone") for a conversational feel.
- Chorus consistency: always repeat the chorus identically unless the user requests variation. Copy-paste, don't paraphrase.
Contrast Between Sections
Verses: longer lines, more unstressed syllables, conversational.
Chorus: shorter lines, more stresses, punchy and direct.
Suno-Specific Anti-patterns
- Lyric bleed in style prompts — the most common mistake (see Style Prompt Formatting Rules above)
- Commas in style prompts — use periods and "and/with" instead (see Style Prompt Formatting Rules above)
- Artist names — Suno rejects them; describe sonic characteristics instead
- Stacking unrelated imagery ("neon skies, electric hearts, endless dreams") — signals AI-generated lyrics; pick one metaphor and develop it
- Paraphrasing the chorus — always copy-paste it exactly
- Padding lines with filler words to hit syllable count — restructure instead
Emphasis and Dynamics
Building Effective Personas
For consistent vocal character across generations, build a character dossier:
Layer 1: Demographics and Timbre
Age, gender, voice type, fundamental character
Layer 2: Technical Delivery
Enunciation, phrasing, breath control, vocal techniques
Layer 3: Emotional Context
Detached, passionate, vulnerable, aggressive
Layer 4: Sonic Anchor
Reference points that give Suno a clear target — describe the sonic quality rather than naming artists
Example:
Female contralto, androgynous, cold, monotone delivery, sharp enunciation, emotionally numb, sinister tone, industrial darkwave atmosphere with crushing bass and high-harmonic distortion.
Use the persona consistently across style prompts for a coherent "artist" sound. Suno's Persona Voices feature (V5.5) can lock this in for reuse.
Quality Checklist
Before delivering lyrics:
Prosody Audit
Every song must pass a detailed prosody audit before submission. No summary passes — a "PASS" without receipts is not a pass.
The audit must include:
- Line-by-line syllable counts for every sung line
- Rhyme scheme map per section (AABB, ABAB, etc.) with classification: true rhyme, near-rhyme, or assonance
- Singability check at the target BPM — how many syllables fit per bar? Flag lines that rush or drag
- Genre-appropriate meter: e.g., NDW = 8–12 syllables, punchy, staccato; French chanson = octosyllabic or alexandrine. Match the genre's conventions.
Why this matters: Rhyme fixes break meter. Meter fixes introduce filler. The only reliable gate is a line-by-line audit with visible evidence. If the checker says "PASS" without showing syllable counts, send it back.
Handling User Input
If user provides existing lyrics
- Ask how to adjust (never reproduce copyrighted lyrics verbatim)
- Suggest improvements: meter, rhyme, structure
- Offer to rewrite in the style of, not copy
If user provides a theme/concept
- Ask about genre/mood preferences
- Propose a structure (verse-chorus-verse-bridge-chorus)
- Draft lyrics with proper formatting
- Provide matching style prompt
Style Prompt Examples
genre: "operatic rockabilly, fast-paced, upbeat"
vocal: "male vocals, hymnic chorus"
mood: "energetic, triumphant"
genre: "industrial krautrock with heavy German accent"
instrumentation: "industrial equipment, distortion, motorik beat"
production: "raw, mechanical, relentless"
genre: "French yéyé synthwave, 112 BPM"
vocal: "breathy female vocal, call and response"
instrumentation: "tenor saxophone, gated reverb, handclaps"
mood: "anthemic chorus, nostalgic"
genre: "dark wave rockabilly with NDW influence, 130 BPM"
vocal: "powerful sultry female vocals with punk edge"
instrumentation: "slapped upright bass, moody analog synths"
V5.5 Features
This skill is optimized for Suno V5.5 (March 2026), which produces 48kHz broadcast-quality audio.
Key V5.5 Capabilities
- Persona Voices: Create reusable vocal identities for consistent sound across songs (Pro/Premier)
- Voice Cloning: Clone your own voice with a verification process (Pro/Premier)
- Custom Models: Train up to 3 personalized models from your song library (Pro/Premier)
- My Taste: Adaptive preference system that learns from your generation history (all users)
- Song Editor: Replace sections, extend, crop, and fade for iterative refinement
- Creative Sliders: Weirdness, Style Influence, and Audio Influence for fine-tuning generation personality
- 12-stem separation: Export individual instrument stems for DAW mixing (Pro/Premier)
Tips for V5.5
- Style prompts can now be up to ~1,000 characters — use the extra space for detailed instrumentation and production notes when needed
- Parameterized metatags (
[Verse: descriptor] syntax) are more reliably followed in V5.5
- Use Persona Voices for consistency when generating multiple songs for the same "artist"
- The Song Editor is more cost-efficient than regenerating entire songs — fix weak sections instead
Suno API (via suno CLI)
The suno shell command drives Suno through the user's own browser session on suno.com. It is a convenience wrapper around the same web API the suno.com UI uses.
Prerequisites — user actions, not agent actions
Before you can call suno, the user must:
- Open https://suno.com in their own browser.
- Sign in with their Suno account.
- Keep the tab open while the skill is in use.
The agent must never bypass these steps, prompt the user for a password, or attempt to log in on the user's behalf.
How authentication works
The script runs fetch() calls inside the suno.com page via playwright-cli eval. The session token is read from the page and consumed by fetch() in the same evaluation block — it is never printed, stored, copied to the agent's context, or written to disk. Only the API response payload leaves the browser.
If the user is not logged in, the script exits with a clear error directing them to sign in.
Credit safety
suno generate and suno extend spend real Suno credits on every call. Always:
- Present the final lyrics, style prompt, sliders, and persona to the user.
- Wait for explicit user approval in the conversation.
- Then invoke the command.
Read-only commands (credits, feed, personas, voices, poll, search, clip, playlists, projects, me) are safe to run without confirmation.
Quick start
suno credits
suno personas
suno generate --lyrics "[Verse]\nHello world" --tags "rock, indie" --title "My Song"
suno generate --lyrics "..." --tags "..." --title "..." --persona <persona-id>
suno generate --simple "a funky disco track about robots"
suno feed
suno poll <clip_id> --wait
suno search "indie rock" --type=public_song
All commands
generate, poll, feed, clip, search, lyrics, lyrics-status, trash, credits, rename, visibility, extend, tags, playlists, projects, personas, voices, me
Run suno help for full usage, or see references/endpoints.md for the complete API reference.
Submitting Songs via UI Automation
For UI-based submission (as fallback or when the API isn't available), use playwright-cli to automate the Create page. Navigate to Create, select Custom mode, optionally choose a persona, fill in lyrics and styles, set the title, configure advanced options, and click Create.
For the complete step-by-step guide, see references/suno-ui-automation.md.
Full Song Examples
See references/examples.md for complete lyrics with formatting.
Reference
For detailed metatag lists and advanced techniques, see references/metatags.md
For detailed realism and synthesis descriptor vocabulary, see references/descriptors.md