Your revised prompt
Optimizing your prompt…This may take a moment.
Free online tool
Prompt Optimizer
Turn a rough prompt into clear instructions. Compare the changes, then copy the revised prompt into your AI assistant.
No signup required Copy-ready output Save in this browser
Start here
Improve your prompt
Ready
Your revised prompt
PreviewGive your draft a clearer direction.
Paste your prompt and select Optimize My Prompt. Your revised wording will appear here when processing finishes.
How it works
What is a prompt optimizer?
A prompt optimizer revises instructions you already have. SecondBrain uses your draft and optional goal, context, must-keeps and output format to produce a candidate prompt. You can inspect the text changes, copy the result and save it locally.
Keep the details that matter
Include the facts, constraints and commitments your answer depends on. Check that the revision preserves them; extra words or a shorter prompt do not establish an improvement.
Check the actual answer
Run the prompt in your chosen assistant. Compare the answer against your requirements, including factual accuracy, completeness and format.
The web tool produces prompt text. It does not run your task, inspect a private account or test the answer for you. The optional Chrome extension offers a separate workflow inside supported chatbots.
How SecondBrain does prompt optimization
From a rough prompt to a reusable draft
Use your own requirements to guide a revision, inspect the text changes and save a prompt after checking it.
SecondBrain produces a revised prompt from the instructions and context you supply. Review the draft and test it in your chosen assistant.
1
Paste the rough prompt
Start with the messy, quick prompt you already wrote. Skip learning complex new frameworks or mastering prompt engineering acronyms.
2
Clarify the goal and tighten the wording
Add your goal, must-keeps and format when they affect the task. The requested format is an instruction to the model, not schema enforcement.
3
Compare the text changes
The comparison shows added and removed lines and word counts. It does not score quality or establish why an edit will improve an answer.
4
Save the prompts that work
Turn one successful result into a repeatable asset. Build a personal prompt library directly in your browser to eliminate repetitive typing across recurring workflows.
5
Use the prompt in your assistant
Copy the prompt into ChatGPT, Claude, Gemini or your chosen assistant. Supply any referenced files there and check the answer for that model and task.
Use the prompt generator to start from a task description. Use this optimizer to revise instructions you already have. For a planning framework, explore the marketing-plan prompt; for repeatable work tasks, browse ChatGPT prompt templates.
Prompt revisions by use case
Make the requirements explicit, then check whether the answer actually meets them.
These revisions are authored illustrations, not recorded web-optimizer runs or performance comparisons. Each uses the same supplied facts and requirements on both sides.
An update that keeps an uncertain date uncertain
Original instructions
Write an email for fictional Bramble Works stakeholders, 120–170 words with a subject line. The inventory pilot targets 3 November 2026; the date is not confirmed. One integration test remains, with a review on 20 October. Ask them to report blockers to Mara Chen by 19 October. Use a calm, professional tone and do not invent progress or test results.
Authored revision
Draft a ready-to-send project status email for stakeholders at fictional Bramble Works. Facts to preserve: 3 November 2026 is the inventory-pilot target, not a confirmed date. One integration test remains; its review is on 20 October. Action: Ask stakeholders to report blockers to Mara Chen by 19 October. Format: Subject line and 120–170-word email; calm, professional tone. Use only these facts. Do not imply the test passed or invent progress, results or commitments.
The revision groups the existing facts, action and format. Check the dates, target qualification and blocker request in the final email.
A coding request with a defined missing-value behavior
Original instructions
Update the supplied Python function and input example. When user_id is missing, return None; keep all other behavior. Return only the corrected function. If the supplied code is incomplete, ask for the missing code. [Paste function and input example here.]
Authored revision
Task: Update the Python function supplied below. Required behavior: Return None when user_id is missing. Preserve all other behavior. Output: The corrected function only. Use the supplied function and input example. If the code is incomplete, ask for what is missing rather than inventing it. [Paste function and input example here.]
The same missing-value behavior and output requirement appear on both sides. Test present and missing user_id cases against the actual code.
A summary that preserves proposals and open questions
Original instructions
Summarize the supplied project notes in a table with Item, Status and Source phrase columns. Keep proposed dates and budgets conditional. Include unanswered questions. Do not add outside facts. [Paste project notes here.]
Authored revision
Summarize only the project notes supplied below. Return a table with Item, Status and Source phrase columns, followed by unanswered questions. Preserve qualifications: a proposed date is not a confirmed date; a proposed budget is not an approval. Use an exact phrase from the notes to support each row. Do not add outside facts. [Paste project notes here.]
The revision separates the requested format from the source boundaries. Check each table row against its source phrase.
Check the answer
How to evaluate the revised prompt
Compare answers against the same requirements. Keep the prompt that serves the task; length alone is not a quality measure.
Choose your criteria before comparing answers. Give each approach the same facts and context, use the same model where practical, and check more than one example.
Pick the success metric before you edit
You need a scorecard before you need a rewrite. Define task success based on your operational requirements.
- Writing: Tone match, structural adherence, and CTA inclusion.
- Coding: Test pass rate, error resolution, and exact syntax correctness.
- Extraction: Valid JSON syntax, exact field match, and zero hallucinations.
Test against real-world inputs
Use the same supplied facts, requirements and reference inputs for both prompts. Compare several answers, including an easy case and a difficult one; record regressions as well as improvements.
Stop when gains flatten
Avoid endless prompt tweaking. Over-optimization leads to diminishing returns and risks overfitting the instruction to a single edge case. Once the output quality stabilizes and passes your rubric, save the prompt and move on.
Use the same checks for both answers
| Check | What to look for |
|---|---|
| Facts and commitments | Names, numbers, dates and uncertainty match the supplied information. |
| Task completion | The answer delivers the requested work and includes every required action. |
| Output format | The structure, length and fields meet your requirements. |
| Regressions | The revision does not lose something the original handled correctly. |
For a fair comparison, give both approaches the same facts and context. You can include a direct request to the assistant as another baseline.
See a recorded summary example with its source notes and model output. That gallery template run illustrates the checks; it is not a before-and-after test of this web optimizer.
When Prompt Optimization Helps Most
A revision is worth testing when missing instructions, repeated work or specific output requirements make the answer harder to get right.
High-value, recurring workflows
For recurring communications, extraction or domain-specific tasks, a checked reusable prompt can reduce repeated setup. Verify its results before using it for consequential work.
Model migration
When changing a model, test your existing prompt on representative inputs. Revise the instructions only where you find a specific failure or unmet requirement.
When to skip it
If your prompt already works, there may be little reason to rewrite it. Direct prompting can also be enough for a simple one-off task.
What the tool changes
Can I use this as a prompt enhancer?
Yes. Use your existing prompt, then add the requirements the revision should preserve. Whether you call it enhancing or optimizing, review the changes against the task you actually need to complete.
Clarify the instructions
Supply the intended result, relevant facts, output format and things to avoid. Review the revised prompt for missing requirements, invented details and unnecessary instructions.
Check the resulting answer
SecondBrain returns a candidate rewrite and a text comparison. It does not run both prompts or score their answers. Test the revision in your chosen assistant using the same facts and checks as the original.
You can also rewrite a prompt yourself or ask an assistant to do it. Compare approaches using the same information and judge the resulting answers against your task.
Saved Prompt Library
Keep prompts you can reuse
Keep useful revisions for tasks you repeat. Saved prompts stay in this browser.
Saved in this browser
0 saved prompts
Frequently Asked Questions
Get clear answers on how SecondBrain works, where it lives, and how it handles your specific workflow.
What is a prompt optimizer?
A prompt optimizer revises existing instructions using the goal, context and requirements you provide. SecondBrain returns a candidate prompt and a text comparison. Review the revision, then test the answer in your chosen assistant.
What is the difference between a prompt optimizer and a prompt enhancer?
The labels overlap and do not define a standard set of features. SecondBrain revises an existing prompt using your requirements and shows text changes. It does not run a before-and-after answer evaluation. Test the resulting answer; longer or shorter wording alone does not establish an improvement.
Can prompt optimization make prompts shorter?
It can remove repetition, but it may also add instructions your task needs. Check factual fidelity and task completion before treating a reduction in length as an improvement.
Does a prompt optimizer work with ChatGPT, Claude, and Gemini?
You can copy the revised prompt into ChatGPT, Claude, Gemini or another assistant. Supply required files or context there, and evaluate the answer for that model and task.
How do I know the optimized prompt is better?
Give the original and revised prompts the same facts and context, use the same model where practical, and compare several answers against a rubric chosen in advance. Check regressions as well as improvements.
Related prompt tools