Anthropic's K-12 Skills Are the Strongest New Domain
The strongest new domain in the batch is K-12 education. The four Anthropic teacher skills average 9.07, higher than any repository in the batch.
| Skill | Trigger | Structure | Workflow | Content | Engineering | Security | Total |
|---|---|---|---|---|---|---|---|
| k12-check-for-understanding | 9.4 | 8.9 | 9.2 | 9.1 | 8.8 | 9.7 | 9.2 |
| k12-lesson-differentiation | 9.5 | 8.9 | 9.4 | 9.2 | 8.8 | 9.8 | 9.3 |
| k12-lesson-plan-creation | 9.0 | 9.0 | 9.0 | 9.5 | 8.5 | 9.5 | 9.1 |
| k12-lesson-prep | 9.0 | 7.5 | 8.5 | 9.0 | 8.5 | 9.5 | 8.7 |
Across all six education skills in the batch, including two generic lesson-planning skills from other repositories, the mean is 8.92.
What the four share
An activation test. Each skill says what the teacher should have in hand before it starts: a lesson objective, a grade band, a topic, a class context. The agent is not asked to guess.
A concrete output. The files specify what the teacher receives: a lesson plan with timed segments, a differentiated set of tasks, a set of checking questions with expected answers. The output is not "advice".
A differentiation axis. The strongest file in the set, k12-lesson-differentiation, does not just say "differentiate". It names the axes along which a lesson can be adapted and the constraints that must remain fixed.
A check for understanding. The k12-check-for-understanding skill is a full procedure for producing questions and the expected student responses, not a prompt to "ask some questions".
Not just the brand
Anthropic's name is not why the scores are high. The generic lesson-plan from genspark scores 8.6, which is also strong; it differs mainly in trigger precision and the specificity of its output format. The weaker generic files in the batch show that "education" is not a magic category.
The K-12 set is a good template for any domain where the output has to be usable by a human immediately: state the inputs, name the pedagogical constraint, produce a concrete artifact, and include a way to check that the artifact works.
What is still missing
Four skills do not make an education category. We still have no coverage of assessment design beyond check-for-understanding, no curriculum mapping, no parent communication, and no higher-education workflows. But this is the strongest start in the batch, and the average is not carried by one outlier: all four are above 8.5.