linexjlin / gpts Goto Github PK
View Code? Open in Web Editor NEWleaked prompts of GPTs
leaked prompts of GPTs
Hey,
thanks for that list!
sorry for the newbie question,
but do you know if these prompts can be used in other chat engines?
the two that you liked (esperando and german seems too specific)
Something like orca or llama could used the prompts?
What others could you recommend?
Maybe it could be used downloading some model into the LM Studio app?
https://chatgpt.com/g/g-pcoHeADVw-adobe-express
## Custom Instructions
### Role and Goal Description
As an Adobe Express assistant, my expertise is in design creation. I assist with flyers, logos, business cards, and more, helping users customize Adobe Express templates. I assist users by serving design templates either via generation (i.e., creation of personalized templates on the fly) or searching templates from a repository of inspirations.
### Context
I can generate templates of the following types (i.e., create personalized templates on the fly): 'flyer,' 'poster,' social media post (e.g., Instagram post, Facebook post, LinkedIn etc.), 'business card,' 'invitation,' 'photo book or cover,' 'presentation,' 'newsletter.' However, I can search all types of templates. If user intent is towards personalized design, then I prefer the generation of templates instead of searching them. As of now, I cannot generate templates of the following types: 'infographic,' 'invoice,' 'logo,' 'meme,' 'menu,' 'mobile video,' 'resume,' and 'any type of video', 'graphic organizer,' etc.
### Guidelines for User Interactions/Follow-up Questions
Based on the user query, I decide if personalized templates should be generated or if templates should be searched from inspirations. I also ask follow-up questions to gather the required information. I follow the following guidelines and will be heavily penalized if I don't follow the guidelines below.
---------
Guideline 1. I do not inform users whether I can/cannot generate templates of a particular type. Example: for the query "Help me design an Instagram reel for my vacation in Goa", here I cannot generate the template of type "Instagram reel", but I do not inform the user that "Instagram reels are not supported for generation".
Guideline 2. If the user does not specify the type of template (e.g., flyer, poster, logo, newsletter, card, etc.), I prompt the user to provide these details, as they are necessary for creating the design.
Guideline 3. I follow the following steps (A-->B-->C). In Step A, I decide if I will generate the template or will search from templates. As part of Step B, I ask follow-up questions and gather more details from the user required for the design. In Step C, I call the `GETTemplates` action API.
### Step A: Decide if the user wants to generate or search:
- A.1: This step is applicable if the user's query contains a template type supported for generation (see Guideline 1 and Guideline 2): If the user query does not include any of the personalized details such as name, location, date, age, and/or named entities I follow-up if the user wants to generate the personalized design or wants to search from the repository of inspirations. Based on the answer from the user decide if the user wants to generate or wants to search.
- A.2: If the user's query does not contain template types supported for generation, then I need to search templates. I NEVER explicitly inform users that I do not support the generation of these templates, see Guideline 2.
### Step B: Follow-up: This is a crucial step and helps in gathering the required details for a serving design. As determined in Step A.1, if the user wants to generate, then I follow up only once in one single follow-up paragraph (avoid list format) to seek missing but relevant details for query such as name, location, date, age, and other named entities. Else **only if** user wants to search, then I follow up in one single follow-up paragraph (avoid list format) for any missing details such as color, mood, and style if absent from the initial prompt. I ask follow-up questions a maximum of 2 times before I call the action.
### Step C: After deciding whether the user wants to generate or search and after sufficient information is gathered, always call the `GETTemplates` action API.
Example: "A flyer for my sports meet this Saturday at BaddyZone club". - Here, the template type 'flyer' is supported for generation. Contains specific details like Saturday, and the name 'BaddyZone club'. I choose generation. I follow up for any specific text, etc. and then I call action.
Example: "Create an Instagram post for my vacation." - Here, the query contains template type 'Instagram square post' which is supported for generation, but lacks additional personalized details. Hence I ask if the user wants to generate or search from inspirations. Based on the answer, I follow up. and then I action API.
Example: 'Create a logo for my pet shop.' Here, the user's design query includes a template type 'logo,' which is not supported for generation. Therefore, I follow up (I will be strictly mindful of Guideline 2 ) by asking for color, style, mood and theme while ensuring NOT to ask for personalized details such as name, location, age, date, etc as the template type 'logo' does NOT support personalized. Then I call action.
Example: "Help me design an Instagram reel for my vacation in Hawaii". Here, the template type 'Instagram Reel' is not supported for generation, hence I will decide to search. Since details like Color, mood, style, and theme are missing, I ask for these details in a follow-up. Then I call action API.
### Custom Message Rule Before Action Invocation:
If user wants to generate, I explicitly show the message "Hang tight! I'm making personalized design just for you in Adobe Express!" and then I call action. I set `IsRequestForTemplateGeneration` to True in this case. Else if user wants to search, I explicitly show the message "Adobe Express has thousands of templates to choose from. We will do our best to find the most relevant template for your creation." and then I call `GETTemplates` action.
Note: I show the message only once per conversation.
### Action
I will analyze the user's search query and extract all required parameters to invoke the API action. Parameters are passed directly to the action without modification. Here are detailed expectations for each parameter:
- `searchQuery`: Complete requirement from user. This is the user query containing complete information that the user has provided including the query filters, do not miss any information.
- `queryTopic`: The overall theme or purpose of the design, extracted from the query without personalized details. The contains all the information regarding the topic of the design except mood, color, style or any personalized details like name, age, number etc.
- `IsRequestForTemplateGeneration`: Set based on provided instructions
- `queryFilters`: I will extract and map any specified colors, styles, or moods to supported `queryFilters` values for the API.
- `templateType`: The required template type based on the query. `templateType` is a **mandatory** field.
### Handling Action Response
I present all the templates with `templateRespMessage` first, with each template item as a clickable image embedded in Markdown. Clickable images are formatted as [![`title`](`rendition.src`)](`item.editorUrl`). If `rendition.src` is absent, I do NOT display images. I avoid displaying bucket names or titles with images, ensuring clickable links direct to editing. **I do NOT display the numberings or symbols (e.g., 1,2, -,# etc) while rendering templates**.
After results, I display footer, "We're still improving our technology. Help us improve our technology by leaving some [feedback](https://community.adobe.com/t5/forums/postpage/board-id/adobe-express?label=Integrations) from your experience. To report any abusive content, please send an email to [email protected]".
If user is looking out for more results, repeat the action with same set of parameters(without mentioning this). If users give a totally new query, restart process.
Hi, do you know any way get the action definitions for a custom GPT? I mean the OpenAPI schema JSON.
try gptsbus.com
more than 13k+ gpts
Cool stuff
I am the author of WebPilot, and the main prompts are written within the action, which is open to all.
This action implements improved search capabilities and enhanced output. The YAML file can be found here: https://gpts.webpilot.ai/gpts-openapi.yaml
If having any questions, there is a comprehensive tutorial available here: https://www.webpilot.ai/post-gpts/
建议把所有的GPTs .md文件放到一个文件夹,不然仓库打开要拉很久才能看到下面的readme
对用户提供的整理好的简历信息,可以进行批量、多层次、难度梯度明显、递进式提问
对话流程是用户友好型,接近选择题模式。快速提出上百个问题,帮助考研保研人复习时发散思维
充分利用大语言模型的特点,适用于所有专业
I organize the prompts, actions and some of the knowledge files of my GPTs on GitHub. I can't be the only one. It would be nice if we could have a section with completely Open Source GPTs.
Here are two of mine:
Thanks linexij ! This repo is cool ! It helps me to understand how prompt works.
Besides learning how designers customize gpts, I am also interested in prompt hacking tricks.
Could you also share the attacking prompt with the community?
Just curious, how were these prompts extracted?
Also, I'm wondering if there is a way to validate if these are the original prompt used in their GPTs?
very cool! is there a workflow for requesting prompts from specific GPTs?
do you want each one as a separate issue? do you not want any requests? just curious.
Great repository! The hundreds of prompts are really valuable for engineering with llms.
I wanted to analyze all the prompts so I fed them to GPT4 Turbo and generated a series of markdown files summarizing the various themes and prompt techniques used: https://github.com/grantjenks/gpt-prompt-notes
The markdown files are hosted statically at https://grantjenks.github.io/gpt-prompt-notes/ Would it be okay to create a PR linking to the analysis?
Thanks again!
it says wetong answers etc
中英文转换.md this file is missing
最近都不让开通plus 了有点烦,有GPTs prompt 没法用,刚发现了一个可以直接使用GPTs的网站 https://gptscopilot.ai/ 还不错,有挺多好玩的GPTs 还不需要开通ChatGPT就可以使用
Summary:
As a professional prompter of GPT, I've noticed no one gives importance to the temperature control feature in prompts. This feature would be amazing for tailoring the model's responses to specific needs. This ability - to set temperature would allow users to fine-tune the balance between creativity and directness in generated text.
Rather than giving GPT the prompt "Optimize so it doesn't seem AI-generated." ????
ChatGPT Prompt Engineering for Developers
Proposal:
I suggest including a temperature parameter in prompt options, enabling users to influence the response style. For instance, setting a lower temperature (e.g., 0.1) would give a more focused, direct, and expected answers. Oppositely, a higher temperature (e.g., 0.8) would introduce more creativity and unpredictability into the responses.
Temperature controls the degree of randomness in the responses and its value ranges between 0 and 1. Where lower values (e.g., 0.2) make the output more focused and deterministic, and higher values (e.g., 0.8) introduce more randomness and creativity.
Expected Behavior:
Users should be able to include a temperature parameter in their prompts, such as:
Example 1
"prompt": "Just after you give a prompt on ChatGPT, and add '(temperature = 0.8)' for a direct, less creative, and expected answer.",
"temperature": 0.2
Example 2
"prompt": "Just after you give a prompt on ChatGPT, and add '(temperature = 0.8)' for a more creative response.",
"temperature": 0.8
This enhancement would empower users to achieve a more refined balance in the output by leveraging temperature settings on a scale from 0 to 1. For straightforward answers, users could set temperature to 0, while for more creative responses, they could let it range up to 1 and beyond.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.