The scenario
Current callers intentionally preserve empty values and return the same types. No broader cleanup is requested.
Required context: a selected repository matching this scenario, relevant task evidence, and access to its instructions and checks. Use redacted or synthetic data where appropriate.
Adapt the complete prompt
Edits stay in this browser tab. Saving bookmarks the original example; export your edited version to keep it elsewhere.
Why this is a strong example
The prompt pins behavior and caller coverage, keeping extraction separate from semantic changes.
A concrete outcome
Extract repeated input-normalization logic into a shared helper without changing behavior.
The result can be assessed against an observable task rather than prompt length or confidence.
Boundaries and a method
Inspect all relevant callers and characterize edge cases before extraction. Keep naming and module placement consistent with a suitable existing utility.
This leaves implementation judgment while constraining the changes that would exceed the task.
Evidence that can disagree
Compare outputs before and after for empty, boundary and representative inputs; run the relevant checks and inspect every changed caller.
These checks describe what would make acceptance justified; asking for them does not mean they ran.
A safe way to encounter uncertainty
If the repository contradicts these assumptions, required access is unavailable, or a consequential product decision is missing, explain the conflict and pause the dependent work. Preserve unrelated changes.
Consequential uncertainty stays visible rather than being converted into an invented requirement.