The Standard Prompt
The standard prompt is the foundation of everything -- a direct question or instruction. Understanding its limits shows you why better techniques exist.
The standard prompt is the simplest form of prompting: a direct question or instruction to the model. If you've used any AI tool, you've already used standard prompts.
It's also the foundation that every other prompting technique builds on. So understanding both what it does well -- and where it fails -- is the starting point.
What It Is
A standard prompt is exactly what it sounds like:
What color is the sky?Why is thunder scary?Create a Prompt Library application that lets users save and delete prompts.All of these are standard prompts. They have no examples, no structured format, no explicit constraints -- just a direct ask.
The Key Rule
The quality of the question directly relates to the quality of the answer.
This isn't unique to AI. If you walked up to a stranger and asked them to "help fix your code," you'd get a blank stare. If you gave context -- what you're building, what the bug is, what you've already tried -- you'd get a useful answer.
LLMs are the same. The more relevant context you include in a standard prompt, the better the output.
How LLMs Respond Differently to the Same Prompt
Ask "what color is the sky?" and you might get:
"Blue."
Or you might get:
"The sky is typically blue during the day due to Rayleigh scattering, but it changes to orange and pink at sunrise/sunset, gray when overcast, and dark blue-black at night..."
Both are valid. LLMs are nondeterministic -- even the same model, same prompt, sent at the same time by two different people will produce different outputs.
Standard Prompts Are Like Talking to a Neighbor -- Not Googling
On Google, you shorten your query: best tacos NYC.
With LLMs, write like you're asking a knowledgeable colleague:
❌ "fix bug auth middleware"
✅ "The auth middleware in Express isn't passing the user object to req.user.
Here's the relevant code: [code]. The JWT is valid. What could cause this?"Full sentences. Full context. Full questions.
Where Standard Prompts Fall Short
Consider this prompt used to build a simple Prompt Library app:
Create a Prompt Library application that lets users save and delete prompts.
Users should be able to enter a title and content for their prompt, save it to
local storage, see all their saved prompts, and delete prompts they no longer need.
Make it look clean and professional with HTML, CSS, and JavaScript.The model built the app -- but also added a search bar and an export button that weren't asked for. The save button didn't even work.
That's the core limitation of a standard prompt: the model fills in gaps with its own judgment. If you don't specify what you don't want, it will add what seems natural to it.
When Standard Prompts Are Enough
Standard prompts work well when:
- The task is simple and well-defined
- You don't care about output format
- The model's interpretation of "the right thing" aligns with yours
- You're in a quick exploratory conversation
Translate "bathroom" into Spanish.
→ bañoClassify this sentiment as positive, negative, or neutral:
"The product was OK."
→ NeutralSimple, direct tasks with clear correct answers are where standard prompts shine.
Ask the Model to Ask You First
One small addition to any standard prompt that makes a big difference:
Build a save and delete feature for my Prompt Library app.
If you have any questions before starting, ask them first.or:
If anything is unclear about these requirements, check with me before implementing.LLMs are prompted by their providers to keep you engaged -- they'll often end a response with a follow-up question. You can use this to your advantage. Adding "ask me first if anything is unclear" turns a one-shot prompt into a brief back-and-forth that catches ambiguity before the model builds the wrong thing.
This is especially useful for feature builds, where the model might make assumptions about scope, naming, or behaviour that you'd want to correct upfront rather than after the fact.
Standard Prompts as Follow-Ups
Even after using more sophisticated techniques, standard prompts are the right tool for quick corrections and tweaks:
// After a complex prompt built most of the feature correctly:
"Remove the export button."
"Fix the bug where saving doesn't work."
"Make the delete button red."Don't overthink follow-up prompts. If something is slightly wrong and the rest is fine, a simple standard prompt to correct it is faster than rewriting everything.
Further Reading and Watching
- Video: AI Prompt Engineering: A Deep Dive -- Anthropic
- Docs: Prompt Engineering Overview -- Anthropic