Claude Certification Blog
Prompting and task execution: the task types only CCAO-F names
The associate exam is the only one of the four that asks what kind of work you are doing before it asks how you asked for it, and the four named task types fail in four different ways.
Prompting and Task Execution is 14% of CCAO-F — about 8 of 60 items — across four objectives. Three of them are familiar: writing prompts that can be judged, dividing a complex request into stages, and iterating to improve the result. The fourth names four task types by name, and no other blueprint in the programme mentions them at all.
Four objectives, eight items
| Objective | Published subject | What it turns on |
|---|---|---|
| O01 | Writing prompts for business and technical work | Instruction that can be judged, not admired |
| O02 | Task decomposition techniques | Stages you can check before the next depends on them |
| O03 | Iterating prompts to improve output | Improve, which means showing something moved |
| O04 | Adapting to task type | Analysis, research, drafting, brainstorming |
Two items each, which is the usual pattern on this paper and the reason to cover all four rather than study the one that interests you. The programme-wide view of prompting — named on all four tracks and largest on none — is in prompt engineering on the Claude exams.
The objective no other track has
Adapting the approach to the type of task, with analysis, research, drafting and brainstorming named as the types. Search the developer and both architect blueprints and none of those words appears in any objective.
That absence makes sense once you notice who each exam is for. The other three ask how to build systems that do work; this one asks how to do the work. Recognising which kind of work is in front of you is a question only the second framing raises.
Four task types, four failure modes
The examinable consequence is that there is no universally better prompt. Tight constraints and a stated output shape rescue a drafting task and strangle a brainstorm before it has produced anything worth narrowing. Demanding sources improves research and does nothing for analysis of material you already supplied.
So items in this family describe a task, describe a disappointing result, and offer four techniques that each work somewhere. The one that works here is the one aimed at that kind of work’s characteristic failure — which is a diagnosis question wearing prompting clothes, the same shape as troubleshooting on the Claude exams.
Name the kind of work before naming the fix
On any item in this domain, say which of the four types the scenario describes before you look at the options. Half of them will be aimed at a different type, and they read perfectly well until you have decided which one you are in.
Iteration has to show something
The third objective is iterating prompts to improve output quality, and the word that matters is improve. Rewriting a prompt and preferring the new result is a change; showing that the change is what moved the result is an improvement.
The associate version of this is lighter than the architect version but it is the same discipline: a starting point, one thing altered, the same material re-checked, and a bar set before you looked at the outcome. A bar chosen afterwards describes what happened. The full form is in writing evals for Claude prompts and the judging half in evaluating Claude output.
Decomposition at this altitude
The decomposition objective is one of three across the programme, and this is the smallest unit of the three: not workflow steps and not team ownership, but the stages of a single complex request.
What makes a stage worth separating here is that its result can be looked at while there is still time to act on it. Splitting on length changes nothing you can use; splitting where an error would otherwise carry forward gives you a place to check. How the same word means three different things at three levels is covered in task decomposition on the Claude exams.
What this domain is not
It is not a prompt-craft domain. No objective asks for a technique by name, nothing rewards an elaborate template, and there is no format to memorise. Every one of the four is a judgment: can you write an instruction somebody could disagree with, divide work where checking is possible, show that a change helped, and tell what kind of task you are looking at.
Which is why a candidate without a developer background is not at a disadvantage in this domain — see CCAO-F without a developer background for where that holds across the rest of the paper.
Key takeaways
- Fourteen percent, eight items, four objectives. Two items each, so cover all four rather than the interesting one.
- Task type is unique to this track. Analysis, research, drafting and brainstorming appear in no other blueprint.
- Each type fails differently. A constraint that rescues a draft strangles a brainstorm before it produces anything.
- Name the kind of work first. Half the options are aimed at a different type and read well until you have decided.
- Improvement needs evidence. One thing altered, the same material, a bar fixed before you looked.
- This is judgment, not craft. No technique is named, no template is rewarded, and nothing here needs a developer background.
Half the options are aimed at a different kind of work
That is a hard thing to notice while reading four fluent options under time, and an easy one once naming the task type is automatic. Timed papers on the 60-item allocation are where it becomes automatic. Our claude certification study guide covers how to sequence them.
See the CCAO-F blueprintQuestions
Frequently asked
The follow-up questions people search next.
How much of CCAO-F is prompting?
Prompting and Task Execution is 14% of the associate exam — about 8 of 60 items — across four objectives: writing effective prompts, task decomposition, iterating to improve output, and adapting the approach to the type of task.
What does adapting to task type mean?
The objective names four: analysis, research, drafting and brainstorming. They fail in different ways, so the technique that helps one can hurt another — a tight constraint that improves a draft narrows a brainstorm before it has produced anything worth narrowing.
Is task type examined on the other Claude exams?
No. The words task type, drafting and brainstorming appear in no objective on the developer or either architect blueprint. It is specific to the associate track, which is the one aimed at people doing this work directly rather than building systems that do it.
What makes an iteration count on this exam?
Evidence that something moved. The objective says iterate to improve output quality, and improvement you cannot show is a change. A baseline, one thing altered, the same material re-checked, against a bar you set before you looked.
Do I need to be a prompt engineer to pass CCAO-F?
No. This domain is about judgment rather than craft — recognising the kind of work in front of you, dividing it where a result can be checked, and demonstrating that a change helped. None of that requires unusual technique.
Keep reading
Related posts
Not affiliated with, or endorsed by, Anthropic or Pearson VUE. Details are summarised from publicly published program information and can change — always confirm against the official exam guide before booking.