2 posts were split to a new topic: Request: Start new chat on ChatGPT website
This is my version of ClipboardGPT:
try it please.
Hello,
Do you think it would be possible to add conditions?
I often use ChatGPT to correct or improve my messages.
However, today I have to add the conditions, correct, improve, etc…
So, it forces me to go back to the top of the message and add this condition.
But also to remove it afterwards.
It would be very convenient if we could add our conditions in the extension.
Thank you for your feedback.
Bonjour,
Pensez-vous qu’il serait possible d’ajouter des conditions ?
J’utilise fréquemment chatgpt pour corriger ou améliorer mes messages.
Cependant, aujourd’hui, je dois ajouter les conditions, corriger, améliorer, etc …
Par conséquent, ça m’oblige à revenir en haut du message et ajouter cette condition.
Mais aussi à la supprimer par la suite.
Alors que si on pouvait ajouter nos conditions dans l’extension, ça serait très pratique.
Merci de votre retour.
1 Like
Adding the ability to wrap selected text inside a predefined prompt, with multiple prompts available and easy UX to select them, would be a powerful feature for this extension.
1 Like
Hey Nick, Would you add a checkbox to the ChatGPT Extention so that copy to clipboard could be set as the default to eliminate the need to hold the shift key each time?
Wow, better than what I asked for. Thanks!
1 Like
You can now add conditions in the extension settings, in the System Message box.
Hello everyone,
I’m encountering a small issue with the plugin.
I’ve added the condition “improve:”, which is the same one I use with the plugin that works in the application.
However, when I select a text (which I mostly do to improve and correct emails), the plugin replies to my email instead of improving it.
It’s worth noting that when I use the plugin with the application, I don’t experience any issues.
I’m guessing the plugin isn’t correctly taking my condition into account?
It does take note of the system message for me. It could be a prompt design issue. I would suggest a wordier system message like “Please improve spelling and grammar, and lightly copyedit the following text”
I’ll give it a try, thanks for your feedback!
But it’s odd that it works perfectly with one plugin and not the other.
The difference may lie in using a system message via the API versus including the prompt in the message text itself. A system message typically requires a more explicit “statement of purpose,” such as “Act as a copy editor, correcting spelling and grammar in the supplied texts. Reply with only the corrected text and no additional comments.” This is not quite the same as prepending the message with an instruction like “improve:”.
o1’s opinion on this matter after I sent it a screenshot of this thread:
Thought about image analysis for 18 seconds
It looks like the core issue is that the PopClip extension which relies on ChatGPT is handling your “improve:” instruction differently depending on whether it’s used as a system message in the API or simply as an inline/user prompt. In other words:
- When you use the “improve:” keyword in one plugin, it may be placing that instruction explicitly into a system‐level message. Because system messages have more authority in ChatGPT’s hierarchy of instructions, ChatGPT follows them closely (e.g., it really does only “improve” the text).
- When you use it in the other plugin, it’s possible that “improve:” is just being treated as part of the user prompt rather than a higher‐priority system instruction—or is being appended in a way that ChatGPT interprets differently. Hence, ChatGPT might respond with a more general reply to your text instead of purely returning an improved copy.
Nick’s suggestion to use a more explicit system‐level instruction—something like:
Please act as a copy editor. Improve the spelling, grammar, and clarity of the following text. Respond only with the corrected text, and do not add extra commentary.
is meant to ensure that ChatGPT is in “edit/copyedit mode” by default. This sort of system directive tends to produce much more consistent results than a short prompt like “improve:” in some contexts, especially when using ChatGPT through an API or plugin.
Why more explicit wording helps
- Clearer instructions let ChatGPT know the exact role it should take, such as “copy editor” or “proofreader.”
- Specifying the response format (“respond only with corrected text…”) leaves less room for ChatGPT to assume a more general or conversational style.
- Ensuring it’s a system or role message can override ChatGPT’s default behavior more reliably, so it won’t break into a more standard “chatty” reply.
Why one plugin might behave differently than another
- The successful plugin may be inserting your instructions into a system‐message slot or be implementing them in a single structured prompt that’s recognized as a request for an edit.
- The other plugin might be putting your “improve:” instruction in a user message or adding it to the conversation in a way that the model interprets as context for a normal conversation, instead of a direct copyediting command.
If you have the ability to edit how the second plugin forms its requests (e.g., how it arranges prompts in system vs. user messages), try adjusting it so that “Please improve the following text” (or similar) is clearly placed in the system instruction. You should see more consistent “improve/copyedit” outputs that way.
Thank you for the detailed response 