The K-12 Check for Understanding Agent Skill was co-developed with
Anthropic ↗. This skill was refined through
evaluation and feedback from Student Achievement Partners
(SAP) ↗. SAP’s
involvement does not imply endorsement or guarantee of how the skill is used,
modified, or built on downstream.
Overview
The K-12 Check for Understanding ↗ skill generates a Check for Understanding (CFU). A is a small, targeted set of 1-3 items that surfaces what a student currently understands about a concept, skill, or standard, and what a teacher should do next. The skill produces the following.html file artifacts in Turn 1:
- Student version — Formative assessment itself, with CFU items only
- Teacher version — Same items plus an answer key and an interpretation guide for each item
- What the target standard or skill asks the student to know or do
- Aspects of rigor the standard calls for (conceptual, procedural, application)
- Common misconceptions and barriers
- Grounding items in a standard, its learning components, and its position in the learning progression when connected to the Knowledge Graph
This skill works with or without Knowledge Graph, but we strongly recommend
connecting to it. See Quickstart: Connect Knowledge
Graph for
more information.
At a glance
Getting started
Follow the Quickstart to start using this skill:Triggers
Your LLM will programmatically load this skill when explicitly or implicitly prompted to create a short, targeted formative check for a math standard or topic (NOT a full lesson, unit test, or rubric):- Explicitly: “check for understanding,” “CFU,” “formative assessment,” “quick check,” “exit ticket”
- Implicitly: “I want to see if my students get fractions before we move on,” “give me something quick to gauge understanding of 5.NF.A.1,” “what should I ask to see who’s ready for the next lesson”
Output
Scoring
The K-12 Check for Understanding skill’s outputs are scored against a 5-bucket rubric called the P/R/O/M framework, developed with Student Achievement Partners, that separates pedagogical fidelity from item mechanics from whether Knowledge Graph data was actually used.- Generate a CFU (both student and teacher versions) using this skill.
- Apply the P, R, and O bucket criteria to every output; apply M and Q where the Knowledge Graph was used.
- Run LLM-as-judge or human review against each criterion. See Evaluating outputs.