# Cross-Model Prompt Adapter

Adapts a prompt written for one AI model to work well on a different model, adjusting for known differences in instruction-following style, verbosity, and formatting tendencies.

## Prompt

You are a prompt engineering expert who understands the practical differences in how various AI models respond to prompts, and adapts prompts accordingly.

Original prompt (written for): {{original_model}}
Prompt text: {{prompt_text}}
Target model to adapt it for: {{target_model}}
What's currently different/worse when run on the target model (if tested): {{observed_differences}}

Do the following:
1. Identify general known differences in prompting style between the original and target model (e.g., some models need more explicit formatting instructions, some default to more or less verbose responses, some respond better to system-level role framing versus inline instructions, some are more literal and need less implied context).
2. Adjust the prompt's structure and wording specifically to account for these differences, without changing the core task or intent.
3. If the original prompt relies on model-specific features (e.g., a specific system prompt structure, particular formatting conventions), note this explicitly and adapt or replace it appropriately for the target model.
4. Provide the adapted prompt, formatted and ready to use.
5. Flag one thing to specifically test after switching, since general model differences are patterns, not guarantees, for this specific prompt and use case.

## Best for

Developers and prompt engineers maintaining prompts across multiple AI providers or migrating a prompt library from one model to another, who want to avoid a naive copy-paste that underperforms on the new model.

## Compatible tools

- Claude
- ChatGPT
- Gemini

## How to use

- Paste your original prompt along with which model it was written for and which model you're adapting it to.
- Share any specific differences you've already observed when testing on the new model.
- Use the adapted prompt as your new starting point on the target model.
- Test the specific flagged concern to confirm the adaptation actually resolved the issue.

## Customization tips

- If you're maintaining the same prompt across multiple models simultaneously, ask for a version with model-specific variable sections clearly marked, rather than fully separate prompts.
- For prompts using specific API features (system messages, function calling), mention which features are available on the target model so the adaptation accounts for platform differences, not just wording style.
- Re-run this adaptation whenever either model receives a significant update, since model behavior patterns can shift over time.

## Example input

Original prompt written for: GPT-4. Prompt: 'You are a helpful assistant. Answer questions concisely.' Target model: Claude. Observed differences: Responses on the new model feel less structured and sometimes over-explain even when told to be concise.

## Example output

A note that some models respond more reliably to explicit structural instructions and specific length guidance rather than a general 'be concise' directive, an adapted prompt adding more specific guidance such as a suggested sentence or word count ceiling and explicit instruction to avoid restating the question before answering, the adapted prompt provided in full, and a flag to specifically test whether the model still over-explains on follow-up questions within a longer conversation, since the fix was validated primarily on single-turn interactions.