From da2ba07edd3b77fa251e616fe0d13c7f052e89fe Mon Sep 17 00:00:00 2001 From: rmurrills Date: Fri, 5 Dec 2025 22:38:58 +0000 Subject: [PATCH] Add General prompt advice --- General prompt advice | 169 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 169 insertions(+) create mode 100644 General prompt advice diff --git a/General prompt advice b/General prompt advice new file mode 100644 index 0000000..429512d --- /dev/null +++ b/General prompt advice @@ -0,0 +1,169 @@ +# Effective AI Prompts: A Practical Guide + +## Core Principles + +### 1. Be Specific and Clear +- State exactly what you want, **not what you don't want** +- Include relevant constraints, format requirements, and scope +- Avoid ambiguity that could lead to multiple interpretations + +**Poor**: "Tell me about dogs" +**Better**: "Write a 200-word summary of the top 3 health considerations for senior Labrador Retrievers" + +### 2. Provide Context +- Give background information relevant to the task +- Specify your knowledge level or intended audience +- Include why you need this information if it affects the approach + +**Poor**: "Explain quantum computing" +**Better**: "I'm a software developer with no physics background. Explain quantum computing in terms of how it differs from classical computing for solving optimization problems" + +### 3. Define the Output Format +- Specify structure (bullet points, paragraphs, table, code, etc.) +- Set length expectations (word count, number of items) +- Request specific elements to include or exclude + +**Poor**: "Give me marketing ideas" +**Better**: "Provide 5 marketing ideas for a local coffee shop, formatted as: [Idea name] - [Brief description] - [Estimated cost: low/medium/high]" + +## Advanced Techniques + +### Chain of Thought +Ask the AI to work through problems step-by-step rather than jumping to conclusions. + +**Example**: "Solve this problem step-by-step, showing your reasoning at each stage: [problem]" + +### Role Assignment +Give the AI a specific role or perspective to adopt. + +**Example**: "You are an experienced project manager reviewing this timeline. What risks do you see?" + +### Few-Shot Examples +Provide 2-3 examples of the input/output pattern you want. + +**Example**: +``` +Convert these casual phrases to professional email language: +"thanks a bunch" → "Thank you for your assistance" +"let me know" → "Please advise at your convenience" +"ASAP" → "at your earliest convenience" + +Now convert: "can you check this out?" +``` + +### Constraints and Boundaries +Explicitly state limitations or requirements. + +**Example**: "Suggest 3 dinner recipes that: are vegetarian, take under 30 minutes, use no more than 8 ingredients, and don't require an oven" + +## Common Mistakes to Avoid + +### 1. Vague or Open-Ended Requests +**Avoid**: "Help me with my code" +**Instead**: "This Python function is throwing a KeyError on line 42 when the input dictionary is missing the 'name' key. How should I handle this gracefully?" + +### 2. Assuming Context You Haven't Provided +**Avoid**: "What should I do next?" (after describing a problem without your goals) +**Instead**: "Given that my goal is to reduce page load time to under 2 seconds, and I've already optimized images, what should I prioritize next?" + +### 3. Overloading a Single Prompt +**Avoid**: Asking for 10 different unrelated things in one prompt +**Instead**: Break complex requests into sequential prompts, building on previous responses + +### 4. Not Specifying Expertise Level +**Avoid**: "Explain machine learning" +**Instead**: "I'm a beginner programmer. Explain what machine learning is using simple analogies without mathematical notation" + +## Prompt Patterns + +### For Analysis +"Analyze [subject] focusing on [specific aspects]. Identify [what to look for] and explain [what insights you need]." + +### For Creation +"Create a [type of content] about [topic] that [specific requirements]. It should include [elements] and be [tone/style]." + +### For Problem-Solving +"I'm trying to [goal] but encountering [problem]. I've already tried [attempts]. What alternative approaches would you suggest?" + +### For Comparison +"Compare [A] and [B] in terms of [criteria 1], [criteria 2], and [criteria 3]. Present as a table with pros and cons." + +### For Review/Critique +"Review this [content type] for [specific aspects like clarity, accuracy, tone]. Provide specific suggestions for improvement." + +## Iterative Refinement + +Good prompting is often iterative: + +1. **Start with a clear but basic prompt** +2. **Review the output** - what's missing or incorrect? +3. **Refine your prompt** with more specific instructions +4. **Add constraints** based on what went wrong +5. **Provide examples** if the format isn't right + +**Example progression**: +- First: "Write a product description for a coffee maker" +- Second: "Write a 100-word product description for a premium automatic coffee maker, highlighting ease of use and quality" +- Third: "Write a 100-word product description for a premium automatic coffee maker. Start with a compelling benefit, include 3 key features (ease of use, programmable settings, quality brewing), and end with a call to action. Tone should be enthusiastic but professional." + +## Special Considerations + +### For Code Generation +- Specify the programming language and version +- Mention any frameworks, libraries, or dependencies +- State coding standards or style preferences +- Include error handling requirements +- Specify if you need comments or documentation + +**Example**: "Write a Python 3.11 function using type hints that validates email addresses with regex. Include error handling for invalid inputs and docstring documentation. Follow PEP 8 style guidelines." + +### For Creative Writing +- Specify genre, tone, and style +- Define target audience +- Set word count or structure +- Mention any themes or elements to include/avoid + +**Example**: "Write a 300-word science fiction short story opening for young adults. Tone should be mysterious and intriguing. Include a protagonist discovering something unexpected. Avoid graphic violence." + +### For Data and Research +- Specify what timeframe or sources are relevant +- Request citations or sources if needed +- State if you need current information vs general knowledge +- Define the level of technical detail required + +## Quick Checklist + +Before submitting a prompt, verify: +- [ ] Have I clearly stated what I want? +- [ ] Have I provided necessary context? +- [ ] Have I specified the desired output format? +- [ ] Have I set appropriate constraints (length, style, scope)? +- [ ] Is my prompt specific enough to avoid ambiguity? +- [ ] Have I indicated my knowledge level if relevant? +- [ ] Am I asking for one clear thing, or should I break this into multiple prompts? + +## Examples: Poor vs Good + +### Example 1: Technical Help +**Poor**: "My website is slow" +**Good**: "My React website's homepage takes 8 seconds to load. Chrome DevTools shows the main bottleneck is a 3MB bundle.js file. What's the best approach to reduce this bundle size?" + +### Example 2: Content Creation +**Poor**: "Write about climate change" +**Good**: "Write a 500-word article explaining the connection between climate change and extreme weather events for a general audience. Include 3 specific examples from the past year and conclude with actionable steps individuals can take. Tone should be informative but not alarmist." + +### Example 3: Learning +**Poor**: "Teach me Spanish" +**Good**: "I'm planning a trip to Spain in 3 months and know zero Spanish. Teach me 10 essential phrases for ordering food at restaurants, with pronunciation guides and cultural tips on when to use them." + +### Example 4: Decision Making +**Poor**: "Should I use React or Vue?" +**Good**: "I'm building a dashboard application for internal business users that needs real-time data updates, will be maintained by a team of 3 developers, and needs to integrate with our existing REST API. Compare React and Vue for this specific use case, considering learning curve, ecosystem, and real-time capabilities." + +## Remember + +- **Garbage in, garbage out**: The quality of AI responses is directly proportional to the quality of your prompts +- **Experiment**: Don't be afraid to try different phrasings or approaches +- **Iterate**: Refine based on what you get back +- **Be conversational**: You can build on previous responses and ask follow-up questions +- **Leverage AI strengths**: Pattern recognition, summarization, explanation, and transformation tasks work particularly well