The Bloom Taxonomy Coach skill for Claude

The problem the skill solves is named in the file itself, and it is recognisable to anyone who has ever written a test. Many tests measure only the bottom two levels: remembering and understanding. And many learning objectives start with grasp or know, words you cannot observe. You cannot see whether someone grasps something. You can see whether someone explains it, applies it, compares it, evaluates it or designs it. With Bloom, every objective gets a verb you can actually see and assess, and you also cover analysis, judgement and creation.

It is written for teachers, trainers, course creators, tutors and parents helping with homework, from primary school to corporate training. Prior knowledge of Bloom is not needed: the levels and the verbs sit in the file. It is part of the skill library NL we make freely available through the AI and automation service, without an account and without a sales email afterwards. New to skills? What Claude skills are NL first explains exactly what you are downloading.

One thing you should know upfront: the skill organises, it knows nothing about your subject. It knows the taxonomy, not your teaching method. What it delivers is a starting point the subject teacher checks, and that is stated in the file too. If it is unsure about a term, it writes that down instead of guessing.

It starts with the verb. A learning objective that starts with the learner understands or the learner knows gives you nothing to look at. You cannot observe understanding, only behaviour that points to understanding. That is why Anderson and Krathwohl turned the levels into verbs in 2001. From that moment on, every level stated something a learner does, and the objective automatically became testable.

The top half costs the most time. Coming up with a good analysis question or evaluation task takes longer than writing ten fact questions. That is exactly where it gets stuck. It takes over that work: it delivers a learning objective per level and at least one test question that actually measures that objective, including an answer key or assessment criteria for open questions.

Differentiation is the first casualty. An easy and a challenging variant per level, a rubric to go with it, a quiz for the whole class: it is on everyone's wish list and it rarely happens. In the skill that is not a separate track but an option in the same run. You ask for it and it comes along.

And the language has to scale. The same material for Year 4 and for a corporate training needs different sentences, different examples and different question forms. So language and complexity scale with the audience you specify. That is also exactly why the audience sits on the mandatory input checklist: without that information it cannot choose.

In diff form, showing the difference the skill makes to a learning objective about the water cycle. Left what usually stands, right what you can observe.

Then comes the theory the skill rests on, written out in the file itself: Bloom published the Taxonomy of Educational Objectives in 1956 with a committee of examiners, and Anderson and Krathwohl revised the model in 2001. Two changes are essential for the skill. The levels became verbs instead of nouns, so you can observe and test them. And the top two swapped places, putting creating at the top.

The core of the file is a list of eight fixed actions. It asks for topic, audience and desired format. It divides the material across all six levels and states per level what the learner does there with this material. It writes an objective per level in the fixed notation, with a verb from that level's repertoire, drawing on more than thirty verbs per level. It writes at least one test question per level that actually measures the objective, including an answer key or assessment criteria for open questions.

It checks every question on four points and rewrites it if one does not hold up. It delivers differentiation on request in the same run. It scales language and complexity with the audience, from Year 4 to professionals in a corporate training. And it flags subject matter where doubt exists with a check note for the teacher, instead of guessing.

Before anything gets written, the skill runs through a mandatory input checklist: the topic and, if available, the source material, the audience with age or education level and relevant prior knowledge, the desired format, which levels take priority if not all six are needed, and the practical constraints such as number of questions, available test time and open or closed questions. It only asks about what is missing, so you get no form but at most two or three targeted questions.

The output follows a fixed order of five parts: an overview of topic, audience and format plus the assumptions you need to check, the six objectives per level, the test questions per level with the expected answer or the assessment criteria and a time indication, the differentiation if you asked for it, and finally the check note for the teacher. That last one is the most underrated block: it states which subject matter needs checking, which questions lean hardest on prior knowledge, and where the set is thin. Want to learn to write files like this yourself? Writing a SKILL.md NL explains the structure.

The ordering comes from Bloom and his committee. The Taxonomy of Educational Objectives, Handbook I on the cognitive domain, appeared in 1956, written by a committee of examiners edited by Benjamin Bloom. The purpose was practical: examiners wanted a shared language to describe what a test question actually asks. Those six categories are still the backbone of the skill.

The revision comes from Anderson and Krathwohl. A Taxonomy for Learning, Teaching, and Assessing appeared in 2001, with the two changes that make this skill usable: verbs instead of nouns, and creating at the top. Krathwohl also wrote an overview article about that revision in Theory Into Practice in 2002, which is in the source list too. By default it uses the revised version, with the fixed rule that the old and the new are never mixed up without saying so.

The notation comes from Mager. Preparing Instructional Objectives from 1962 is about exactly the point where most learning objectives fail: an objective must describe what the learner can do afterwards, not what they know. That is the source of the fixed notation after this the learner can [verb] [object]. Anyone wanting to learn to formulate objectives more broadly naturally ends up next at the SMART Goals Coach skill.

The coherence comes from Biggs and Tang. Teaching for Quality Learning at University from 2011 describes constructive alignment: learning objective, learning activity and test have to fit together. An objective at the analysing level paired with a test full of fact questions is not just sloppy, it steers your learners towards the wrong kind of learning. That is why it keeps those three together: an objective per level, and the question that measures that objective right alongside it.

Even without installing the skill you can improve your own test with these principles: start every objective with a verb you can see happen, check whether your question actually asks what the label promises, and write your answer key before you decide the question is good.

After that, using it is simple: name your topic, your audience and what you want. If you get stuck unzipping or finding the right folder, that is explained step by step in installing Claude skills NL. And if you want the broader explanation of working with AI first, you find that in the knowledge base.

There are also situations where you are better off leaving it alone. Official attainment targets and exam specifications are one of them: those are set in law, and the skill refuses to rewrite or replace them. Assessment of individual learners is a second: the skill writes questions and criteria, but passes no judgement on a child and invents no grades or norms. And the third limit is the subject matter itself. It knows the taxonomy, not your method. What exactly is in the chapter and what you call the terms, you know and it does not.

One more honest limit the file states itself: the skill cannot guarantee that a question measures the claimed level for every learner. An application question becomes a remembering question the moment the learner already knows the example. That does not make the question useless, it means the teacher knows what they are looking at. Anyone who wants to train that kind of thinking about their own understanding further can turn to the Feynman Technique Coach skill, which keeps simplifying material until the gaps become visible.

A set of objectives with matching questions is a small document with big consequences: it decides where your lesson is working towards and what your learners are going to study for. So start there, and pick up the rest afterwards. Want your team to learn to set up this kind of work themselves with AI? That is exactly what we do in the AI training NL. And if you want to talk through what AI can take over elsewhere in your organisation, from separate skills to full automation, we simply do that in a conversation.