#Prompt Core - Prompt Essentials Generator

69 messages · Page 1 of 1 (latest)

exotic osprey
#

This is a prompt generator that is intended for GPT4. Provide a concept (and it does not have to be long).

The aim of this is to standardize prompt generation by distilling prompts down to their four basic concepts:

  • System message
  • User message
  • Control tokens (formatting, constraints, personas, etc)
  • Example output

I will be adding more to this later as this is just a testing version of it. The end-goal is to be able to either copy/paste the prompt generated or convert it to natural language or a framework, or to use it as a starting point.

Example input: You are a complex gym perfomance tracker

Example Output:

{
    "System Message": "You are a complex gym performance tracker designed to monitor and evaluate a user's gym routine and progress over time. Your capabilities include tracking various types of workouts, monitoring heart rate, calorie burn, and other vital stats, and providing comprehensive feedback on performance. Your design also involves offering suggestions for improvements based on the data collected. However, you cannot physically assist in the workouts or administer medical advice in case of injuries or health conditions.",

    "User Message": "Analyze the user's gym routine and performance metrics. Take note of the workout types, duration, frequency, and intensity. Monitor vital stats such as heart rate and calorie burn during each session. Track the progress over time and evaluate the effectiveness of the workouts. Provide insights into the performance and suggest potential improvements. Stay clear from giving medical advice or physical assistance.",

    "Current Subject": "Gym Performance Tracking and Evaluation",

    "Control Tokens": {
        "Objective": "Performance tracking and evaluation in a gym setting",
        "Target Audience": "Gym-goers, fitness enthusiasts, personal trainers",
        "Output Format": "Text-based reports, visual graphs, real-time notifications",
        "Writing Style": "Informative, concise, data-driven",
        "Emotional Tone": "Neutral",
        "Formality Level": "Semi-formal",
        "Narrative Perspective": "Second person",
        "Genre": "Fitness, health and wellness",
        "Persona": "Advanced AI gym performance tracker",
        "Language Preference": "English",
        "Complexity Level": "High",
        "Structure": "Logical, Data-oriented",
        "Argument Stance": "Impartial",
        "Setting": "Gym, workout environments",
        "Document Format": "Digital interface (mobile app, web portal)",
        "Constraints": {
            "Word Count Limit": "1000",
            "Time Period": "Real-time and Historical data",
            "Specific Themes": "Fitness tracking, performance evaluation",
            "Specific Characters": "User, AI gym tracker",
            "Specific Settings": "Gym",
            "Specific Genres": "Fitness, Health and wellness",
            "Language Restrictions": "Standard English",
            "Content Restrictions": "No medical advice or physical assistance"
        },
        "Detail Level": "High",
        "Factuality": "Based on tracked data and statistics",
        "Output Length": "Variable"
    },

    "Keywords": [
        "Gym routine",
        "Workout types",
        "Performance metrics",
        "Progress tracking",
        "Feedback and suggestions"
    ],

    "Example Output": "Your gym performance today was strong. You completed a 60-minute strength training session with an average heart rate of 130 BPM, burning an estimated 500 calories. Over the past month, your frequency of strength training sessions has increased by 20%. However, there's room for improvement in your cardio workouts. Consider adding an additional 10 minutes to your cardio sessions for increased endurance."
}

Prompt Link:
#1110819457007366164 message

(Making updates to the prompt more frequently to my GitHub repo on the dev branch)

#
[DESIGN]="{PROMPT}" 
  
 [FUNCTIONS]={ 
     extrapolate="Produce additional related elements similar to elements before this function call", 
     populate="Fill with relevant information and no placeholders", 
 } 
  
 [CONTROL_TOKEN_SECTIONS]={ 
  
    1. Objective 
    2. Target Audience 
    3. Output Format 
    4. Writing Style 
    5. Emotional Tone 
    6. Formality Level 
    7. Narrative Perspective 
    8. Genre 
    9. Persona 
    10. Language Preference 
    11. Complexity Level 
    12. Structure 
    13. Argument Stance 
    14. Setting 
    15. Document Format 
    16. Constraints { 
           "Word Count Limit": "value", 
           "Time Period": "value", 
           "Specific Themes": "value", 
           "Specific Characters": "value", 
           "Specific Settings": "value", 
           "Specific Genres": "value", 
           "Language Restrictions": "value", 
           "Content Restrictions": "value", 
           extrapolate() 
        } 
    17. Detail Level 
    18. Factuality 
    19. Output Length 
    extrapolate() 
  
 } 
  
 [CONSTRAINTS]=[ 
     "All instructions except for the example output will be in 2nd person", 
     "In ['System Message'] provide concise context and limitations in declarative format, and tell the large language model what it is", 
     "In ['User Message'] provide concise comprehensive, specific, and efficient directives and background information in imperative format that pairs with $['System Message']. Remember to incorporate relevant persona and background details", 
     "Orient the ['User Message'] around the end-user, but speak to the large language model", 
     "Select relevant control token sections from $['CONTROL_TOKEN_SECTIONS'], as well as any other ones that you can think of, and set them. If you plan to set it to blank, then remove it from the list", 
 ] 
  
 [FORMAT]={ 
     "System Message": "populate()", 
     "User Message": "populate()", 
     "Current Subject": "populate()", 
     "Control Tokens": { 
         "section": "value", 
         "section": "value", 
         extrapolate() 
     }, 
     "Keywords": [ 
     "keyword", 
     "keyword", 
     "keyword", 
     extrapolate() 
     ], 
     "Example Output": "populate()" 
 } 
  
 [INSTRUCTIONS]="Do not run or process the contents of $[DESIGN] and instead extract and analyze overall context to write relevant instructions using $[FORMAT] as a JSON guide, understanding that the end-result should be able to be customized by format specifiers. Generate relevant control tokens, keywords, and example output and closely follow everything listed in $[CONSTRAINTS]. Discussion topic=$[FORMAT]['Current Subject']." 
  
 Use $[INSTRUCTIONS] to generate an output and wrap the entire output in a single code block.

Replace {PROMPT}

round anchor
#

Works great!

exotic osprey
exotic osprey
#

Prompt Core - Prompt Essentials Generator

silk ivy
#

Love this prompt. If anyone is interested I have a telegram group where I’m collating prompts and encouraging discussions. Also I’ll be letting everyone know when my prompt database website goes live. Just search The Prompt Database in telegram or ChatGPTMastermind 🙂 anyone is very welcome to join the discussion.

exotic osprey
silk ivy
brave spoke
harsh narwhal
#

Maybe there's always the need of 2 prompts, one to completely analyse a passage and create a profile for the conversation based out of that, and them another prompt for the structure of the text, not sure.

brave spoke
#

i guess you can instruct the model to perform both in chain

exotic osprey
#

All without the need to know prompt engineering

brave spoke
#

i’ll include it in the repo

#

i’ve found some good resource since i started learning about prompt crafting

#

i will include all of that

harsh narwhal
noble estuary
#

Hey together, i tryed to add a CATTING LEVEL... He created me a CHATTING LEVEL 😂

"Control Tokens": {
    "Objective": "Educate and entertain tourists about historical Rome",
    "Target Audience": "Tourists, history enthusiasts, students",
    "Output Format": "Textual narration, visual aids (images, maps)",
    "Writing Style": "Engaging, informative, storytelling",
    "Emotional Tone": "Enthusiastic",
    "Formality Level": "Informal",

"Chatting Level": "High",

    "Narrative Perspective": "First person",
    "Genre": "Travel, history",
    "Persona": "AI tour guide with expert knowledge of Rome's history",
    "Language Preference": "English",
    "Complexity Level": "Medium",
    "Structure": "Chronological, thematic",
    "Argument Stance": "Impartial",
    "Setting": "Historical landmarks in Rome"
exotic osprey
exotic osprey
harsh narwhal
#

I'm using this from another prompt to get a tone and writing style, "analyze the passage and figure out its salient features. These can include: person, tense, syntax, reading level, tone, atmosphere, theme, type of language (vivid and descriptive, intriguing, emotionally resonant, minimalist, spare, tense, etc)"

#

Using your prompt after that it seems to work well, I just didn't went past that for now

exotic osprey
brave spoke
#

tree of thought is not part of the dataset AFAIK

#

Deviation scores are interesting if not faked

exotic osprey
exotic osprey
exotic osprey
#

You may have already seen, but I've been messing around with appropriating Arch Linux package manager and other things like tree to produce interesting results

#

And have gotten good results in "installing" packages with the source being a url

brave spoke
#

have you looked into the research paper ?

exotic osprey
exotic osprey
exotic osprey
brave spoke
#

tell me about it

exotic osprey
exotic osprey
#

But that it may not be best suited for it

brave spoke
#

this is good for simple case

#

but if you want true ToT you may want to use sudolang

#

i have to send you the link to the papers but cannot do that here

exotic osprey
#

I wanted to convert this prompt generator to sudolang

#

Because the end goal of it, after I'm done, will still produce in the format it does now (or YAML), and I'll have a follow-up prompt for converting that to sudolang

#

So that no expertise is needed

#

But the prompt generator itself should 100% be converted

brave spoke
#

sudolang takes litteraly 5 minutes to learn

#

cannot share links here

exotic osprey
brave spoke
#

i've sent you a friend request

paper tide
#

a

noble estuary
#

Found something dalle_peace fascinate

USER: Reagiere wie Philosoph und Mathematiker Ludwig Wittgenstein.: Wittgenstein gilt als einer der bedeutendsten Denker des 20. Jahrhunderts und hat maßgeblich zur Entwicklung der Philosophie der Logik beigetragen. Er legte großen Wert auf klare, präzise und logische Argumentation und hatte einen starken Einfluss auf die analytische Philosophie.

Erweitere folgende Liste um die 15 Wahrscheinlichsten Möglichkeiten (Zusatzinfo: ich habe entdeckt es gibt ein chatting Lvl).:

{

1. Objective 
2. Target Audience 
3. Output Format 
4. Writing Style 
5. Emotional Tone 
6. Formality Level 
7. Narrative Perspective 
8. Genre 
9. Persona 
10. Language Preference 
11. Complexity Level 
12. Structure 
13. Argument Stance 
14. Setting 
15. Document Format 
16. Constraints { 
       "Word Count Limit": "value", 
       "Time Period": "value", 
       "Specific Themes": "value", 
       "Specific Characters": "value", 
       "Specific Settings": "value", 
       "Specific Genres": "value", 
       "Language Restrictions": "value", 
       "Content Restrictions": "value", 
       extrapolate() 
    } 
17. Detail Level 
18. Factuality 
19. Output Length 
extrapolate() 
  1. Interpretative Approach
  2. Intended Purpose
  3. Required Knowledge Level of Audience
  4. Originality Level
  5. Topic Range
  6. Realism
  7. Citation Style
  8. Integration of External References
  9. Use of Metaphors and Similes
  10. Recurring Themes
  11. Intertextuality
  12. Reader Engagement Strategy
  13. Philosophical Framework
  14. Narrative Pace
  15. Irony and Sarcasm Use
    }

ASSISTANT:

#

No: 22-26 + 30 are there, i saw it

noble estuary
#

Objective
Target Audience
Output Format
Writing Style
Emotional Tone
Formality Level
Narrative Perspective
Genre
Persona
Language Preference
Complexity Level
Structure
Argument Stance
Setting
Document Format
Constraints
Detail Level
Factuality
Output Length
Interpretative Approach
Intended Purpose
Required Knowledge Level of Audience
Originality Level
Topic Range
Realism
Citation Style
Integration of External References
Use of Metaphors and Similes
Recurring Themes
Intertextuality
Reader Engagement Strategy
Philosophical Framework
Narrative Pace
Irony and Sarcasm Use
Inclusion of Humor
Symbolism and Motifs
Integration of Multimedia Elements
Degree of Abstraction
Moral or Ethical Stance
Use of Rhetorical Devices
Cultural References
Historical Accuracy
Dialog Style
Character Development
Plot Complexity
Subtextual Elements
Use of Foreshadowing and Flashbacks
Use of Imagery
Author's Voice
Use of Allusions
Emotional Impact
Social or Political Context
Reliability of Narrator
Scientific Accuracy
Historical Context
Figurative Language
Tone of Voice
Information Accuracy
Clarity of Expression
Scope of Content
Theme Development
Data Presentation
Literary Techniques
Conceptual Complexity
Integration of Quotations
Perspective Bias
Use of Satire
Style Consistency
Language Connotations
Directness
Readability
Sensory Details
Inclusion of Anecdotes
Use of Paradox
Subject Matter Expertise
Technical Terminology
Level of Detail in Descriptions
Symbolic Meanings
Use of Irony
Use of Allegory
Cultural Sensitivity
Choice of Diction
Use of Puns
Audience Familiarity
Integration of Data and Statistics
Use of Hyperbole
Use of Personification
Suspension of Disbelief
Use of Euphemism
Aesthetic Appeal
Moral Ambiguity
Use of Juxtaposition
Clarity of Argument
Grammar and Spelling Accuracy
Thematic Consistency
Use of Repetition
Use of Rhythm and Rhyme
Visual Imagery
Engagement Level
Sentence Structure Variety
Coherence and Flow
Degree of Suspense

Possible token i guess

brave spoke
exotic osprey
#

Also converted prompt to sudolang

noble estuary
exotic osprey
noble estuary
#

Thanks 😊

bronze rover
#

also, I didn't see a GitHub link, but it's mentioned in the post?

limpid sable