Step by Step
C
Context — providing necessary background
Giving the model the necessary background information it needs to produce a genuinely useful, well-grounded response.
Example: including relevant background facts or a document excerpt before asking a question about it.
L
Length — specifying desired output length
Indicating how long or short the desired response should be, helping the model calibrate its output appropriately.
Example: explicitly asking for "a two-paragraph summary" rather than leaving the desired length ambiguous.
E
Examples — few-shot prompting
Providing 2-5 examples of the desired input-output format (few-shot prompting) to clearly demonstrate exactly what kind of response is wanted.
Example: showing 3 example question-answer pairs in the desired format before asking the model to answer a new, similar question.
A
Ask specifically — precise, unambiguous instructions
Being precise and specific about exactly what's being asked for, rather than leaving the request vague or open to misinterpretation.
Example: asking specifically for "a bulleted list of exactly 5 pros and 5 cons" rather than a vague "tell me about the pros and cons."
R
Role assignment — calibrating vocabulary, depth, and style
Assigning the model a specific role ("Act as an expert...") to calibrate its vocabulary, depth of explanation, and overall style to match the intended context.
Example: prompting "Act as a patient, friendly high school teacher" to get a simpler, more approachable explanation style than a default response might provide.
Applied Walkthrough
1
A user asks an LLM a complex multi-step math word problem and gets an incorrect answer.
2
Ask: what single addition to the prompt often dramatically improves accuracy on this kind of reasoning task? Adding "Let's think step by step" — Chain-of-Thought prompting.
3
This simple addition encourages the model to work through its reasoning explicitly, step by step, rather than jumping straight to a final answer, which measurably improves accuracy on math, logic, and multi-step reasoning tasks.
4
Combined with other CLEAR framework elements — providing clear context, specifying desired length, giving examples, asking specifically, and assigning an appropriate role — this produces a well-engineered prompt likely to get a high-quality, accurate response.
Exam Application
Exams test whether you can name and apply the CLEAR framework elements, and specifically whether you know Chain-of-Thought prompting ("let's think step by step") as a well-documented technique that measurably improves accuracy on reasoning-heavy tasks like math and logic.
⚠ Common Trap
The most common trap is assuming prompt engineering is just about being polite or verbose. Effective prompt engineering is about specific, structured techniques (context, examples, explicit instructions, role assignment, Chain-of-Thought) that measurably and reliably improve output quality — it's a genuine, learnable skill with documented techniques.
✓ Quick Self-Check
1. What does the C in CLEAR stand for?
Context — providing necessary background information.
Tap to reveal / hide
2. What does providing 'Examples' in a prompt refer to?
Few-shot prompting — giving 2-5 examples of the desired input-output format.
Tap to reveal / hide
3. What does Chain-of-Thought prompting specifically add to a prompt?
A phrase like "Let's think step by step," which dramatically improves reasoning accuracy on math, logic, and multi-step tasks.
Tap to reveal / hide
4. What does Role assignment in a prompt calibrate?
The model's vocabulary, depth of explanation, and overall style, by assigning it a specific role like "Act as an expert..."
Tap to reveal / hide
5. What technique specifically helps ground an LLM's response in real facts and reduce hallucination?
RAG (Retrieval-Augmented Generation) — retrieving relevant context to include in the prompt.
Tap to reveal / hide