Prompt Sharpener — your prompt, marked up before your assistant sees it

01Input5/7 filled in

Start

Task

GoalStart here · required

Fine-tune promptSet the environment and point of view

Stack

Role

More fine-tuning optionsThinking, output, scope and constraints

Thinking effort

Output shape

Complexity

Detail

Files in scope

Exclude

Stop condition

Context

Context to includeAdd evidence · required

What makes good context?

Relevant sourcePaste the smallest function, component, query, config, or file excerpt that owns the behavior.
Exact failureInclude the complete error message, stack trace, failing command, or unexpected output.
Expected vs. actualSay what happens now and what should happen instead, using one concrete example.
VerificationAdd the nearest test, reproduction steps, or command that proves the change works.

Keep it focused. Remove secrets, personal data, and unrelated files.

Useful: focused code, exact errors, small logs, configs, schemas, or request/response examples.

Best next addition: Paste the smallest function, component, query, config, or file excerpt that owns the behavior.

02Prompt preview~160 tokens

A focused request, assembled from your inputs.

Start with a clear goal5/7 sections ready · approximately 160 tokens

Goal:

Context:

Scope:- Task type: Debugging - Role: Senior engineer - Stack: Ruby on Rails, PostgreSQL - Complexity: Medium - Response detail: Concise - Selected files or areas: 1

Exclude:Secrets, customer data, generated files, vendored dependencies, full logs, and unrelated modules.

Constraints:- Do not change public APIs. - Do not add new libraries. - Include focused tests for changed behavior. - Use medium thinking effort for normal implementation and debugging.

Output:Return a short diagnosis first, then one minimal patch if the cause is clear.

Stop when:Stop after one minimal patch and a short verification step.

Assistant & costoptional settings

Assistant

Choose where this prompt will run. This selection updates the model and cost estimate above.

GitHub Copilot · Claude Sonnet 4.5