Prompt
The instruction and context you give Claude in a message. A strong business prompt states the role, the context, the task, the constraints, and the format of the deliverable.
Exam context: Expect scenarios where a vague request produces a generic result. The best fix adds specific context and constraints, not stronger words.
See also: 1.1 Effective Business Prompts
Role
A statement of the perspective Claude should take, such as a financial analyst or an HR communications specialist. A role frames tone and expertise, but it does not replace the context and constraints the task needs.
Exam context: A role on its own is rarely the answer. The correct option usually adds audience, source material, or format as well.
See also: 1.1 Effective Business Prompts
Constraints
The limits and requirements on the answer: length, tone, what to include, what to avoid, and what to do when information is missing.
Exam context: Structure beats intensifiers. "Under 150 words, two bullet points per risk" outperforms "be thorough and careful".
See also: 1.1 Effective Business Prompts
XML Tags
Simple tags such as <document> or <instructions> used to separate pasted material from instructions. They are an optional clarity aid, most useful when a prompt mixes several kinds of content.
Exam context: Tags help but are not required. Do not pick an answer that treats tags as mandatory or as a guarantee of accuracy.
See also: 1.1 Effective Business Prompts
Colleague Test
A check on a prompt: if you gave the same text to a capable colleague with no background, could they do the task? If they would need to ask clarifying questions, Claude is missing the same information.
See also: 1.1 Effective Business Prompts
Task Decomposition
Breaking a large or compound request into smaller steps that can be run and checked one at a time. Each step has an output you can verify before the next step depends on it.
Exam context: When a request bundles several jobs and quality is uneven, the best answer is to split it into steps, not to ask for more effort.
See also: 1.2 Task Decomposition
Prompt Chaining
Running a sequence of prompts where the output of one step becomes input to the next. It makes each stage reviewable and lets you correct an early step before it spreads.
Exam context: Chaining fits compound tasks with dependent steps. A self-correction step can review a draft against the source before you use it.
See also: 1.2 Task Decomposition
Self-Correction Step
A later prompt that asks Claude to review its own output against stated criteria or the source. It can catch some problems but is not independent verification.
Exam context: Treat it as a useful second pass, not as proof. Consequential facts still need a check against the source of record.
See also: 1.2 Task Decomposition
Iterative Refinement
Improving an output over several rounds by giving specific feedback: what failed, what to keep, and what to change. It is the opposite of regenerating and hoping for a better draw.
Exam context: The correct response to an unsatisfying draft is targeted feedback tied to success criteria, not repeated regeneration.
See also: 1.3 Iterative Prompt Improvement
Success Criteria
Observable conditions that define a good result, written before or alongside the first prompt. They give you something to score each draft against and tell you when to stop.
See also: 1.3 Iterative Prompt Improvement
Task Type
The kind of work requested: analysis, research, drafting, or brainstorming. Each type calls for different prompt details, such as data and a question for analysis, or audience and voice for drafting.
Exam context: Match the prompt strategy to the task. Brainstorming asks for range first; analysis needs the material provided.
See also: 1.4 Prompt Strategy by Task Type