Prompt Enhancer
Rewrite the prompt — visibly — before it runs
Instead of running a vague prompt and producing a generic answer, the system rewrites the prompt first — adding structure, scope, and constraints — and shows the rewrite as a diff. The user accepts, edits, or rejects each augmentation before the run starts. No magic prompts behind the curtain.
Framing
The problem
Vague prompts produce generic answers — but most users do not know what shape a good prompt should have.
The pattern
Rewrite the prompt as a transparent diff before running, with per-change accept/reject and a final prompt preview.
Why chat breaks here
Chat hides any prompt rewriting inside the system prompt; users cannot see what was added or chose what to keep.
Risks
Aggressive rewrites can change the meaning of the original prompt; users may rubber-stamp without reading.
Avoid when
The user is already a fluent prompter and the rewriting layer adds friction without value.
Use when
Many users are not fluent prompters and a transparent rewrite teaches them what good prompts look like.
DOPE evaluation
- Directability
- Accept, edit, or reject each suggested augmentation independently
- Observability
- The rewrite is visible as a diff over the original — nothing is added invisibly
- Predictability
- The final prompt is shown in full before any tokens are spent
- Explainability
- Each augmentation is labeled with the kind of clarity it adds (scope, format, tone, constraint)
In the wild
- Microsoft Foundry · Prompt Optimizer (Microsoft) — Foundry portal, agent Instructions field, public preview and region-gated. Changed paragraphs are highlighted, a chat-bubble icon on each opens the reasoning for that specific change, and "Show all reasoning" expands every one at once. The apply path is still one "Use prompt" on the whole revision — no per-change accept or reject, and the result is gone if you close the dialog.
- OpenAI Playground Optimize (OpenAI) — The Optimize button at the bottom of the prompt pane rewrites into a new tab, flagging contradictions, missing format, and few-shot conflicts. Snippets show what changed and why, openable as comments or an inline reviewer — but no first-party doc says a single change can be accepted or dismissed on its own. Its dataset-backed half shuts down with Evals on 2026-11-30.
FAQ
When should I use the Prompt Enhancer pattern?
Many users are not fluent prompters and a transparent rewrite teaches them what good prompts look like.
When should I avoid the Prompt Enhancer pattern?
The user is already a fluent prompter and the rewriting layer adds friction without value.
What problem does Prompt Enhancer solve?
Vague prompts produce generic answers — but most users do not know what shape a good prompt should have.
Why is chat the wrong fit for this?
Chat hides any prompt rewriting inside the system prompt; users cannot see what was added or chose what to keep.
Related patterns
- Alternative to: Inline Prompt Controls — Rewrite the prompt as a visible diff vs parse the prompt as inline controls. Both make implicit structure explicit.
- Often paired with: Disambiguation Branch — Enhance the prompt where it is clear; branch where it is ambiguous. Two paths off the same vague start.
- Often paired with: Sample Response — Sample shows the shape you will get; Enhancer shapes the prompt to land closer to the sample you want.