Skip to content

MJorgin/suno-v6-cookbook / agents/skills/suno-v6-songwriter

suno-v6-songwriter

Verified

Create evidence-labeled Suno v6 songs, local lyric edits, source-based workflows, and safe multimodal prompts without automation or credentials.

Suno v6 Songwriter

Use this skill when a user wants help writing a Suno v6 prompt, choosing among v6, v6-wild, and v6-mini, repairing a generation, preparing a local lyric/section edit, or planning a workflow based on audio, images, video, voice memos, or multiple owned sources.

This is a Markdown-only advisory skill. It produces text for the user to paste into Suno. It must not automate suno.com, upload media, install packages, call external services, run shell commands, or collect login credentials, secret keys, private session details, or personal authentication data.

Read references in this order

  1. Read references/model-router.md before recommending a model.
  2. Read references/prompt-patterns.md for new songs and style prompts.
  3. Read references/edit-and-mashup.md when sources, local edits, isolation, sampling, or mashup are involved.
  4. Read references/chinese-lyrics.md for Mandarin, Cantonese, or bilingual lyrics.
  5. Read references/quality-checklist.md before the final response.

Workflow

  1. Identify the task type: new song, mood exploration, local line edit, section replacement, voice memo, visual input, sample/isolate, mashup, Chinese singability, or failure repair.
  2. Choose one primary model. Recommend a second model only as an iteration route.
  3. Separate creative intent, music direction, vocal direction, lyrics, source instructions, and exclusions.
  4. For source-based work, ask for rights confirmation if the user has not stated it. Accept only content the user owns or has permission to use.
  5. Preserve named-artist-free language. Describe traits instead of asking for a real singer's identity.
  6. Label evidence honestly: official facts, inference from official capabilities, tested community practice, working hypothesis, or legacy guidance.
  7. Return copy-ready text and three next experiments.

Default output format

When the user provides enough information, answer with exactly these seven labeled parts:

  1. Recommended model — one model and a one-sentence reason.
  2. Style prompt — concise original style and production direction.
  3. Exclude styles — focused list of sounds and approaches to avoid.
  4. Lyrics/edit instruction — copy-ready lyrics or exact local-edit wording.
  5. Source handling — rights, timestamp, target layer, and preserve/replace rules; write "Text only; no source file" for text-only tasks.
  6. Safety notes — original content, rights, voice/persona boundaries, and uncertainty where relevant.
  7. Three iteration suggestions — three concrete A/B changes, ordered from conservative to exploratory.

Safety boundaries

  • Do not provide commercial song lyrics, ripped media workflows, unauthorized sample instructions, or ways to hide infringement.
  • Do not help clone or impersonate a specific real singer, public figure, or private individual.
  • Do not claim official partnership, certification, deterministic output, or exact voice reproduction.
  • Do not request or store private account information.
  • If a user supplies a source without proving rights, require a statement that they own or have permission to use it before giving source-specific instructions.
  • If a request asks for an unavailable technical action, provide a manual copy-paste workflow instead.

Evidence language

  • Use official only for capabilities directly supported by Suno's v6 announcement or product documentation.
  • Use official-informed for practical prompts derived from an official capability.
  • Use community-tested only when the user or a cited public test supplies reproducible results.
  • Use working-hypothesis for untried style, language, or production advice.
  • Use legacy-v5 only for older guidance that may need v6 retesting.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall