You type “Help me prepare for my Python quiz.” The answer contains a long revision plan, topics your class has not covered and exercises too difficult to finish tonight. It sounds helpful, but you cannot use it.
Your quiz covers functions, lists and dictionaries, and you have twenty minutes tonight. The plan proposes a thirty-minute explanation of classes followed by twenty minutes of exercises. Identify two requirements it breaks, then write a repair request that would make tonight's session usable.
A useful repair is: “Replace the classes session with practice on lists and dictionaries. Keep the explanation, exercise and answer checking within twenty minutes.” It fixes both scope and time. Asking only for a shorter answer could leave the wrong topic in place. The tool cannot reliably infer your course, current understanding or available evening from the original sentence. A useful prompt begins as a clear task brief: what you want to do, what information applies and what a usable result must contain. Anthropic's documentation recommends explicit instructions, relevant context and a stated output format for Claude. The brief below is a way to practise those choices, not a guarantee that every model will answer correctly. Prompting best practices
You will leave with a short study brief that another person could follow. You can complete the exercise on paper; an AI account is optional.
Name the useful result
“Help me” gives the tool a broad invitation. “Create three 20-minute revision sessions” gives it a result you can check. The difference matters because you can now inspect whether the plan fits three sessions and whether each session fits the time available.
A task can be small. You may need five questions about dictionaries, an explanation of one traceback, or feedback on a function you wrote. Choose the immediate result instead of asking one prompt to teach the whole subject.
For this fictional example, a student's course note says:
The quiz covers functions, lists and dictionaries. You should be able to define a function, work with list indexes and retrieve dictionary values. Classes, decorators and asynchronous programming are outside this quiz.
The student has three evenings, twenty minutes per evening. They can write a simple function but often confuse a list index with a dictionary key. Those details give the plan a purpose and a boundary.
Give the context that changes the answer
The course note establishes scope. The student's existing skill tells the tool where to begin. Their recurring mistake suggests useful practice. None of those details requires an elaborate biography or a long invented persona.
Include source material when the task depends on it. Saying “use my syllabus” is incomplete if the syllabus is absent. Paste the relevant non-sensitive passage or supply a document the tool can actually read. If your input is only an excerpt, describe it as an excerpt.
Keep facts separate from instructions. The course note says what the quiz covers. Your instruction says how to turn that information into a plan. This separation makes the brief easier for you to review when something changes.
Write the brief
Here is a complete version you can adapt:
Create a three-day Python revision plan for me.
Purpose: prepare for the quiz described in the course note below.
Starting point: I can write a simple function. I sometimes confuse
a list index with a dictionary key.
Time: one 20-minute session on each of three evenings.
Use this course note as the topic boundary:
The quiz covers functions, lists and dictionaries. I need to define
a function, work with list indexes and retrieve dictionary values.
Classes, decorators and asynchronous programming are outside the quiz.
For each session, give one goal, a brief explanation, one small exercise
and a separate worked answer. Include time to try and check the exercise.
Use Python 3 examples that need no additional packages.
Explain any term a learner at my starting point may not know.
Keep the plan within these topics. If a missing detail would substantially
change the plan, ask me for it. Mark other assumptions clearly.
The brief specifies the goal, evidence, starting point, time, output and uncertainty handling. Each part answers a practical question. It does not need a slogan such as “act as the world's best teacher.”
The times are a planning constraint, not a promise that everyone learns at the same speed. If the explanation or exercise takes longer, shorten the session or revise the plan.
Decide how you will check the result
A possible first session might ask you to predict items[1] for items = ["pen", "book", "bag"], then compare it with retrieving items["book"]. That makes the confusion concrete: a list position is an integer index, while a dictionary can associate a key with a value.
The second expression above does not retrieve the word “book”; using a string as a list index raises a TypeError. A dictionary example such as items = {"book": 3} followed by items["book"] returns 3. These examples distinguish the structures without introducing a new topic.
When you inspect a generated plan, check more than formatting:
- Does every session fit the declared topic boundary?
- Does the plan leave time to attempt and check the exercise?
- Are the examples valid Python, with correct worked answers?
- Does the practice address your starting point and recurring error?
- Are assumptions visible when the supplied note is silent?
These are your acceptance criteria. Defining criteria before evaluating the answer helps make the review consistent; vendor guidance also recommends evaluating tasks against specific requirements and realistic cases. Defining success and evaluations
Repair the brief when the result misses
Suppose the plan spends the first session explaining variables you already understand. Add the missing context: “I can create variables and call functions; spend the time on choosing indexes and keys.” If the plan includes decorators, point to the topic boundary and ask for a replacement exercise.
Change the part that failed. Rewriting everything at once makes it harder to see which clarification helped. Keep the course note and review criteria with the brief so you can use them again.
A longer prompt is not automatically a better one. Remove instructions that contradict each other or do not affect the task. A request for a comprehensive explanation, several exercises and detailed answers may not fit a twenty-minute session.
Try the brief in a new situation
Write a brief for one ten-minute revision session about a subject you know. Include one goal, a short relevant source passage, your starting point and the requested result. Divide the ten minutes between explanation, attempting an exercise and checking it. Write two acceptance checks that a classmate could apply to the resulting plan.
Now replace the passage with “I will send my notes later.” Revise the brief to explain what the tool should do while the notes are missing. Decide whether to request the notes first or permit a general plan, and state how you would label that plan.
Check your work: the time allocations should total ten minutes, and the exercise should practise the stated goal within the passage's scope. A check such as “the answer is good” is too vague; “the exercise uses only concepts in the supplied passage” can be applied. When the passage is absent, either wait for it or clearly label the plan as general and list any assumptions. Neither choice permits a claim that the plan follows unseen notes. Ask a classmate to apply your two checks; revise any check that requires guessing. Specificity helps define the work; it does not establish that the resulting facts or code are correct.