Start with the decision record that already owns the job.

Before touching a Suno control, bring forward the exact song/version plus the most relevant existing record: Sound Identity v1 / FIND Completion Record for identity testing, the current BUILD record for repeatability/construction, or the current CONTROL diagnosis for repair. Do not rebuild those decisions here.

If your current Suno account/model exposes a feature differently—or does not expose it at all—use the current interface and official platform behavior. The creative job matters more than matching an old button name.

1. Personas — continuity control, not your identity

A Persona can be useful when you have a specific vocal/style characteristic worth testing across contexts. It does not prove that the characteristic is yours, that it will remain stable everywhere, or that it should become the artist identity.

Do not: create arbitrary numbers of Personas, assume a Persona equals a vocalist you control, or keep generating just because consistency is imperfect. Record what carried over and what drifted.

2. Covers / reinterpretation — controlled variation, not automatic progression

Use a Cover or equivalent reinterpretation route only when the stage has a real variation question: Can the protected identity survive a different arrangement, genre-adjacent lane, energy profile, instrumentation or production context?

Change the smallest useful variable. A good baseline may need zero additional Covers. When a test is needed, compare the variation against the existing survivor and record requested change versus uncontrolled platform drift.

3. Replace Section / local editing — usually CONTROL

When one local section is the diagnosed blocker and the surrounding build is already approved, a section-replacement feature can be the smallest credible intervention. That is a CONTROL job, not a general BUILD habit.

Protect the approved strengths, name the defect before editing, change only the affected region where practical, and compare before/after. If the problem is actually lyric meaning, vocal perspective, or core sound identity, route back to the stage that owns that decision instead of editing around it.

4. Exclude Styles / negative direction — an implementation constraint

Use exclusion controls when you can name a recurring off-target behavior that interferes with the current FIND or BUILD test. “No pop” is weaker than naming the audible behavior you are protecting against.

Example: instead of excluding a whole genre by reflex, document that bright four-on-the-floor drums keep replacing the intended half-time pocket. Then test whether the exclusion helps preserve the mapped groove without creating new drift elsewhere.

5. Crop / source extraction — preserve lineage

Cropping can preserve a useful source section, hook, instrumental passage, or reference segment. It is an asset/source operation, not a CP1 stage.

Save the parent track/version, time range, reason for the crop, intended use and any relevant source/rights note. A cropped asset should remain traceable to the version it came from.

One job, one controlled intervention, one conclusion.

1. Name the current CP1 stage and exact musical question.
2. Import the upstream record by reference.
3. Choose the smallest Suno control capable of testing that question.
4. Keep protected variables fixed where practical.
5. Listen for pre-defined success/failure evidence.
6. Record platform drift separately from your creative decision.
7. Stop when the result answers the question. More generations are not automatically better evidence.