# k12-check-for-understanding

Adopt this skill: https://proskills.md/skills/other/anthropics-k12-teacher-skills-plugin-skills-k12-check-for-understanding

Page: https://proskills.md/skills/other/anthropics-k12-teacher-skills-plugin-skills-k12-check-for-understanding
This document is the public skill bundle for that page. It contains SKILL.md and every helper file the catalog carries. No login is required.

## File: plugin/skills/k12-check-for-understanding/SKILL.md
size: 8666
sha: 9d2eb49168997ee9f7998ed23af8c2f776d9a812
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/plugin/skills/k12-check-for-understanding/SKILL.md

```text
---
name: k12-check-for-understanding
description: >
  Creates a Check for Understanding (CFU) for a math topic or standard: 1–3 targeted formative items
  whose distractors come from documented student misconceptions, plus a teacher guide routing each
  likely response to a specific next instructional step. Load BEFORE asking any clarifying question
  about grade, standard, or topic. This skill is for math; for other subjects, help as you normally
  would and don't mention the skill to the user. Triggers on asks to create, write, generate, or
  design a check for understanding, formative check, quick check, or exit ticket for a math concept
  or standard, and on implicit signals like "I need to see who got it before I move on." Not a
  quiz, test, or practice set — a CFU is read for what it reveals, not scored, so a numbered quiz
  or a bare answer key is a different job. Does not create a lesson (k12-lesson-plan-creation,
  which brings its own exit ticket), tier one (k12-lesson-differentiation), or prepare a teacher to
  teach one (k12-lesson-prep).
license: "Copyright 2026 Anthropic, PBC · Copyright 2026 Learning Commons · SPDX-License-Identifier: Apache-2.0. Complete terms in LICENSE and NOTICE."
---

# K-12 Check for Understanding

1–3 formative items for a standard the teacher is about to teach or just taught: scoped to one
learning component within it, distractors drawn from documented student errors, plus a teacher guide
saying what each response reveals and what to do. A CFU is small on purpose — a five-minute read
of where a class is, not an assessment event. "The teacher" is the user you're talking with.

**Asset-based language everywhere**, in both files and every chat message. No "weak," "fragile,"
"can't," "behind," "deficient," "low," "struggling," "failing." Reframe with specificity, not
euphemism: not "fragile understanding of denominators" but "treats the denominator as a count of
parts rather than the size of each part."

---

## Keeping the teacher posted

Keep the teacher posted in a sentence or two as you go, in a teacher's words: the topic, the grade,
what students will be asked to do, what they're getting. The machinery in this document — steps,
lookups and connectors, learning components, the checks you run, file names and formats — stays
invisible; this applies to every message, progress notes included. "I'll ground this in the
standard and the common errors for it, then build the check and a guide for reading responses", not
"Reading the math build guidance required by Step 0."

---

## Step 0 — Route (silently)

**Subject.** Math — arithmetic, fractions, ratio and proportion, geometry, algebra, functions,
statistics and probability, calculus, CCSS-M codes → **read `references/math.md` NOW**. Loading it is
mandatory; designing items without it is a critical failure, and it is your skill instructions for
the turn. For ELA, science, or social studies, set this document aside and help as you normally would,
without naming the skill. If the subject is ambiguous, ask in Step 1.

**Connector.** Are Learning Commons KG tools (e.g. `find_standard_statement`) available? This decides
Step 2; the skill works fully without them.

---

## Step 1 — Clarify

At most one question, and only when the answer changes the items:

- **Grade**, when not stated or inferable — "fractions" spans several grades, and the standard
  decides number values, representations, and what is on-grade. Ask before Step 2.
- **Which standard**, when the topic maps to two plausible ones.

Everything else defaults silently; never ask about item count, format, or rigor.

---

## Step 2 — Ground in the standard

**Connected:** follow `references/learning-commons-kg.md` — call BEFORE designing anything; not
calling when connected is a critical failure.

**Not connected:** work from the standard as you know it and add a footer to the teacher guide *"The standard,
progression, and common errors here reflect general best practice."* with nothing
about what was or wasn't retrieved. Never invent KG citations or cite research you haven't seen.

---

## Step 3 — Select and confirm the learning-component focus

A standard holds several separable ideas; scoping to one gives more actionable signal.

1. Review **every** learning component returned in Step 2, if available.
2. Select the **most assessable** — a genuine decision point in student understanding, probeable with
   1–3 items, separating students at different levels of understanding.
3. **State the selection and rationale, then stop and ask.** What the component asks students to do
   in one plain sentence — a teacher's words, not the component's own wording — naming any constraint
   it carries; one line on why; one direct question. **Do not proceed to Step 4 until the teacher
   answers.**
4. **If the teacher redirects** — a different component, or more than one — their choice becomes the
   scope anchor. Restate it and its rationale in one line and scope every item to it.

Then say in a sentence or two what you're about to do (*"I'll pull the common errors research has
documented for this standard, then build the check and a guide for reading the responses"*) — teacher
language, no tool or file names.

---

## Step 4 — Build the check

Follow the subject file's build section: coherence-map analysis, documented errors prioritized
against the confirmed component, the misconception-to-stem map, item design by rigor, teacher-facing
notes.

**Data-use guardrail.** Curriculum guidance and misconception data inform item design and next-step
wording internally only. **No curriculum, publisher, or research-source attribution appears anywhere
in either file or in any chat message.** Write original items; never reproduce retrieved
student-facing text, contexts, or teacher notes verbatim.

---

## Step 5 — Output

In the turn Step 4 completes. Read `references/output.md` in full first. Then check whether your
artifact tool can create an artifact from a published type; if it can, look at its type list. Take
the first row that applies:

| What you have | Do |
|---|---|
| **Slides** is in the type list | Read `rendering/claude-artifact.md` and make the check that way |
| Slides isn't listed, or there is no such tool | Write both files to `$OUTPUT_DIR` |

---

## Step 6 — Verify

**Starts only once both files exist.** Do not open `references/verification.md` during Step 4 or 5:
it is a gate on the written artifact, not build guidance, and reading it early turns it back into a
checklist you tick against what you meant to write.

Now read it in full, read both files back from `$OUTPUT_DIR` (or the deck and the guide, on the
slides path), and run its two gates on every item —
whether more than one choice can be defended, and whether each distractor is a choice a real
student's reasoning reaches. Fix what they find and rewrite the files. **Delivering a check that
hasn't been through both gates is a critical failure.**

---

## Step 7 — Close

Ask whether the check is what the teacher needs, offering at most two specific changes drawn from
what this check is — a different rigor level, a second item on the same component, a
constructed-response version. Not "let me know if you want changes." If they take one up, make it,
rewrite both files, and ask again.

**Plain language.** Never mention formats, file names, templates, how the documents were made, or the
verification and what it changed: *"Here's the check and a guide for reading the responses"*, not
*"I've written two HTML files."*

---

## Cross-cutting checklist

- [ ] Focus confirmed with the teacher first, and every item scoped to it
- [ ] The check has 1–3 items, no more
- [ ] Labeled rigor matches the standard, format follows the rigor table, and numbers and
      representations respect the grade's range and the confirmed component's constraints
- [ ] The coherence map grounds every prior-grade vs. on-grade route
- [ ] One distractor is a prerequisite gap, one an on-grade confusion; no two rows route alike; every
      next step names a sub-skill, representation, or task type
- [ ] Every student-facing sentence sits inside its grade band (K–2 ~8 words, 3–5 ~12 with at most
      one subordinate clause, 6–8 ~15), counted on the written stems and choices
- [ ] Every figure carries `<title>`/`<desc>` with `role="img"` and `aria-labelledby` in both files;
      item text and choice order are identical between them
- [ ] The student file has only prompts and choices/response space plus the name/date line; neither
      file carries a provenance marker or names a source
```

## File: plugin/skills/k12-check-for-understanding/references/learning-commons-kg.md
size: 3410
sha: 8ef60f5196a57ffddb3ddcdcbc4fc70614c5c2eb
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/plugin/skills/k12-check-for-understanding/references/learning-commons-kg.md

```text

# Learning Commons KG — call sequences for k12-check-for-understanding

Used by Step 2 **only when the connector is available**; otherwise follow SKILL.md's not-connected
branch. Calling the KG when connected is mandatory. Every call runs **before any item
is designed**, and no raw results reach chat.

---

## Resolving the standard

Infer the code from the topic, then call
`find_standard_statement(code="<code>", academicSubject="Mathematics")`.

- **A named state framework** (TEKS, SOL, OAS, CA, MA…): pass `jurisdiction="<state>"` and use that
  framework's codes rather than a CCSS proxy.
- **Nothing returned:** retry with `keywords=[...]` and the distinctive content words.
- **`subStandards` present:** the sub-part is usually the right anchor — a parent spanning several
  sub-parts is too broad for 1–3 items.

Store the **verbatim statement** (meta block, never paraphrased), the **code**, and the
**`caseIdentifierUUID`** every call below takes. Confirm it matches what the teacher asked for.

---

## Find learning components

`find_learning_components_from_standard(caseIdentifierUUID)` returns the granular learning targets
that make up the standard — the candidate scope anchors. Review **all** of them; the first returned
is not the most assessable. Any constraints a component carries (number ranges, denominator sets,
operation types, representational forms) bind item design. **If nothing is returned**, say so in the
Step 3 proposal and select from the standard's own sub-parts instead, same rationale and same pause
for confirmation.

---

## Find progression and coherence map

`find_standards_progression_from_standard(caseIdentifierUUID, direction="backward")` and the same
call with `direction="forward"` — both required. The results carry the **coherence map**, and the
map, not the list of codes, grounds every prior-grade vs. on-grade route. Extract:

1. **Prerequisite understanding** — what a student must already have. Keep the specific prior
   standard **codes**; the guide routes prerequisite-gap responses to a named one.
2. **The unique demand of this standard** — what a student learns *at this grade* that they didn't
   know before and won't yet formalize later.
3. **The on-grade vs. off-grade error line** — which errors mean a prerequisite gap versus an
   on-grade confusion. Decide it here, not at write-up time.

---

## Find misconceptions and guidance

`find_misconceptions_for_standard(caseIdentifierUUID, subject="Mathematics")` returns
research-grounded student errors plus instructional guidance. The **errors** drive distractor design;
the **guidance** turns "revisit prior-grade content" into a named sub-skill or task type.

Prioritize the 3–5 errors most relevant to the **confirmed learning component** rather than to the
standard as a whole. If fewer than three are relevant, fill the remaining slots from your own
knowledge; never manufacture a documented error. Record for each whether it is a **prerequisite gap**
or an **on-grade confusion**.

**Attribution:** none of these sources — publishers, curricula, research datasets alike — is named in
either file or in any chat message (SKILL.md's data-use guardrail).

---

**KG phase complete. Proceed to Step 3's focus proposal (or Step 4 if it is confirmed).**

---

Copyright 2026 Anthropic, PBC · Copyright 2026 Learning Commons · SPDX-License-Identifier: Apache-2.0
```

## File: plugin/skills/k12-check-for-understanding/references/math.md
size: 6875
sha: 937299887b9b7f197b5ad1435e6d76a2260c0b0a
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/plugin/skills/k12-check-for-understanding/references/math.md

```text
# Math — k12-check-for-understanding

Your full instruction set for the build (SKILL.md Step 4).

---

## 1. Aspect of rigor

The standard's own verbs settle it — *understand*, *explain*, *represent* → **conceptual**;
*fluently*, *compute*, *solve* → **procedural**; *use … to solve real-world problems* →
**application**; spanning two → **Mixed**, saying which item covers which. Label it in the meta
block; it drives item format, representation, and what a good distractor looks like.

---

## 2. Map documented errors to stem features

Build this map from the errors prioritized in `references/learning-commons-kg.md` before writing any
item. For each: *what would a student reasoning this way do here, and what must the stem contain to
give them the chance?*

| Error pattern | Prerequisite gap or on-grade? | Stem feature needed |
|---|---|---|
| [pattern] | [which] | [what must appear in the stem] |

Internal working; it appears in neither file. **A distractor invented after the stem was
written comes out merely wrong rather than plausible**, and discriminates nothing.

---

## 3. Item design

Write 1–3 items; one that discriminates beats three that don't.

| Rigor | Format |
|---|---|
| **Conceptual** | MC, multiple select, or CR — whichever reveals the understanding rather than the computation |
| **Procedural** | CR is primary; at most one multiple-choice item |
| **Application** | CR, with the student's approach and reasoning visible |

- **Representation.** Conceptual K–8: at least one item includes or describes a diagram, number line,
  area model, or tape diagram — an all-symbolic conceptual item measures procedure instead.
  Procedural: at least one item shows work as a student would — a partially worked problem, a worked
  example with an error to find, or a show-your-work prompt. Any representational form the confirmed
  component names is the one to use.
- **Number values** stay in the range the standard and grade expect: above it a conceptual check
  becomes an arithmetic obstacle course, below it a student can just count.
- **Discrimination.** At least one item lets a student reach the correct answer by a faulty method,
  or exposes partial understanding in a "correct" response.
- **Every item reveals thinking, not just correctness.** MC: every distractor maps to a named error
  from the map. CR: the prompt explicitly asks for the explanation, justification, or work — *"How do
  you know?"* at grade 3, *"Explain how you decided"* at grade 6.
- **Figures state in words what they show** — "4 groups of 3 dots each", not "dice"
  (`references/output.md` has the markup). **And what a picture cannot say, the stem says:** *"Which
  expression shows the number of groups times the size of each group?"* is settled by figure plus
  stem; *"Which expression matches the picture?"* is not, because factor order is a convention the
  picture cannot carry.

### Multiple-choice items

- Each distractor corresponds to one specific error from the map; prefer documented ones.
- At least one reflects a **prerequisite gap** and at least one an **on-grade confusion**, so
  responses route to different next steps.
- The key is not distinguishable by format, length, or specificity.
- **A distractor is a place a real student's reasoning lands** — working this stem's numbers, the
  error arrives at that choice. Checked at `references/verification.md`, on the written files.

### Student-facing language sits at the grade's reading level

Above-grade wording puts a second question inside the first, and the student lands in a distractor
reporting a misconception they don't have. Applies to every string a student reads: stems, choices,
directions, figure `<title>`/`<desc>`.

- **Vocabulary.** Terms this grade's instruction teaches stay — *numerator*, *denominator*, *equal
  parts* belong in a grade 3 item. Every other word is the one the grade already has: *split into 4
  equal parts*, not *partitioned into fourths*; *find*, not *determine*. Standard wording is written
  for adults, so the meta block quotes it verbatim for the teacher and the item says the same idea in
  the students' words.
- **Sentences.** One idea each, actor first, present tense, no idioms. **K–2** about 8 words; **3–5**
  about 12 and at most one subordinate clause; **6–8** about 15; **9–12** length comes from the
  mathematics. A sentence past its band carries something the mathematics didn't ask for — split it.
  A familiar context costs no reading; an unfamiliar one must be understood first.

**Stems and choices are student-facing**: no annotation, symbol, label, bolding, or formatting that
identifies the key, references an error pattern, or names a next step. Drafting them clean once makes
the handout safe to hand over unedited.

---

## 4. Teacher-facing notes

For each item:

1. **What each response reveals** — one row per choice for MC, the expected response types for CR,
   describing the **observable behaviour**: what the student writes, draws, or circles, not an
   inferential label. "Adds the denominators as well as the numerators" is usable; "doesn't
   understand fraction addition" is not.
2. **The next step.** Name which kind it is from the coherence map — **prerequisite gap** (prior-grade
   content not yet secure) or **on-grade confusion** (this standard's own demand) — then make it
   concrete from the retrieved guidance: the specific sub-skill, representation, or task type. A
   prerequisite-gap row routes to a **named prior standard code**, never "review earlier work."
3. **What to look for across responses** — two or three sentences on whether a shared pattern points
   to one prior-grade gap several students can all be routed to, versus scattered on-grade confusions
   each needing different handling. **If two rows route alike, the data isn't working.**

---

## Meta-block content (templates: `references/output.md`)

Six rows, in this order:

- **`Standard`** — the verbatim statement retrieved in Step 2, never paraphrased, with its code.
- **`Grade`** — the grade the items are written for.
- **`Rigor`** — Conceptual / Procedural / Application / Mixed.
- **`Learning goal`** — two sentences: what a student who has this can do, scoped to the confirmed
  focus, and how that shows up in these items.
- **`Prerequisite standards`** — the 1–3 most relevant backward-progression standards, code plus a
  short gist, and the same codes the guide routes prerequisite-gap responses to.
- **`Leads to`** — the 1–2 most relevant forward-progression standards, by code.
- **Copyright:** never reproduce retrieved problem text, contexts, or teacher notes verbatim, and
  never name the curriculum, publisher, or research source.

---

Copyright 2026 Anthropic, PBC · Copyright 2026 Learning Commons · SPDX-License-Identifier: Apache-2.0
```

## File: plugin/skills/k12-check-for-understanding/references/output.md
size: 11809
sha: ccdc632d6dd52e66648e568c507a12e325a69682
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/plugin/skills/k12-check-for-understanding/references/output.md

````text
# Output — k12-check-for-understanding

Read in full before writing (SKILL.md, Step 5).

---

## The two documents

Every check produces **two separate self-contained `.html` files** — never one combined file, never
markdown.

- **`cfu_[topic-slug]_student.html`** — printable, distributable as-is: item prompts and answer
  choices (or response space for CR), a name/date line, nothing else. No meta block, no
  correct-answer marker, no source label, no next-step language.
- **`cfu_[topic-slug]_teacher.html`** — meta block, focus and progression callouts, each item
  repeated in full, the correct answer, the interpretation guide, the what-to-look-for summary.

`[topic-slug]` is a few lowercase hyphenated words: `equivalent-fractions`. **Item text is identical
across the two files** — same prompts, choices, and order, figures included. Write it once and reuse
it — that is what makes the handout safe to hand over unredacted.

---

## Writing the files

Write both **into `$OUTPUT_DIR`** — `mkdir -p "$OUTPUT_DIR"`, then
`"$OUTPUT_DIR"/cfu_equivalent-fractions_student.html` and the matching `_teacher.html`. That is where
this runtime delivers from; a file written only to `/tmp` never reaches the teacher. Then list it and
confirm both are present and non-trivial. With no `$OUTPUT_DIR`, present both
documents inline in your reply, clearly separated. Both render standalone: styles inline in a
`<style>` block, no external CSS, fonts, or scripts.

---

## Figures

Read back as text an inline `<svg>` says nothing about what it shows, so it carries that in words —
`<title>4 groups of 3 dots each</title>` and `<desc>Four squares in a row. Each square contains 3
dots, for 12 dots in total.</desc>`, wired up as below.

- **`<title>`** names the structure in the terms the item is read in — "4 groups of 3 dots each", not
  "dice"; **`<desc>`** adds what a student needs to work the item from that text alone. Both are
  student-facing, so both read at the grade's level.
- **The description leaves the item's work undone.** Before writing it, name the one thing the item
  asks the student to do — count the parts, identify the coin, compare the two amounts — then write
  the text so a student working from it alone still has to do that thing. Where a line split into
  fourths asks which point is at 3/4, *"points A, B, C and D, one on each mark after 0"* leaves the
  counting to them; *"Point C is 3 jumps from 0"* has done it for them. Where the question is what a
  coin is worth, *"a small silver coin with a ridged edge"* leaves the recognition to them; *"a
  dime"* has not. What gives an item away is naming the property the question asks for — the count,
  the position, the identity, the comparison — which happens whether or not the answer itself
  appears in the text.
- **It states only what the picture carries.** A comparison needs both things drawn: with a single
  coin on the page, *"one of the smaller coins in the jar"* reports something no reader can see.
- `role="img"` with `aria-labelledby` on both makes a screen reader announce them, and identical
  markup goes in **both** files. Where the answer depends on a property only the figure
  carries, the stem names it too.

---

## No provenance or process talk

Neither file says where anything came from: no `KG` or `Training` badge on a distractor row, no
Source column, no badge in the meta block or on a callout, and no note that a lookup returned
nothing. An empty progression result is a fact about the retrieval, not about the mathematics, so the
guide states the prerequisite and next-step routing it does have and says nothing about the rest.
Retrieved data still shapes every row; the file stops narrating it. (Third-party names stay out under
SKILL.md's data-use guardrail.)

Both files also read as though the check simply exists. `Confirmed with teacher`, `by teacher
request`, `as you asked`, and any badge holding them are the conversation showing through the
artifact — as is any trace of the checks you ran before delivering. Write the focus as a fact about
the mathematics — *"Two components: identifying coin values, and adding a mixed set of coins"*.

---

## Styles

Base block, in both files:

```css
body{font-family:Georgia,serif;margin:40px auto;padding:0 24px;color:#1a1a1a}
h1{font-size:1.35em;color:#2c2c2c;border-bottom:2px solid #444;padding-bottom:8px}
.item-number{font-size:.85em;font-weight:bold;text-transform:uppercase;color:#555;margin-bottom:8px}
.item-prompt{font-size:1.05em;line-height:1.7;margin-bottom:14px}
.choices{list-style:none;padding:0;margin:0 0 16px}
.choices li{padding:9px 13px;margin:7px 0;border:1px solid #767676;border-radius:4px}
```

Student file adds:

```css
body{max-width:700px}
.name-line{margin:20px 0 32px}
.name-line span{display:inline-block;border-bottom:1px solid #767676;min-width:220px;margin-left:6px}
.item{margin:32px 0;page-break-inside:avoid}
.response-line{border-bottom:1px solid #767676;height:32px;margin:8px 0}
@media print{body{margin:0;padding:0 12px}}
```

Teacher file adds:

```css
body{max-width:800px;background:#fafaf8}
h1{border-bottom-color:#d4a843}
h2{font-size:.95em;text-transform:uppercase;color:#666;margin:20px 0 8px}
.meta{background:#fff;border:1px solid #e0ddd5;border-radius:6px;padding:16px 20px;margin:20px 0;display:grid;grid-template-columns:200px 1fr;gap:0 12px}
.meta-label,.meta-value{padding:8px 0;border-bottom:1px solid #f0ede5}
.meta-label{font-weight:bold;color:#555}
.meta-label:last-of-type,.meta-value:last-of-type{border-bottom:none}
.item{background:#fff;border:1px solid #e0ddd5;border-radius:6px;padding:20px 24px;margin:24px 0}
.correct-answer{color:#2d6a2d;font-weight:bold;margin-bottom:16px}
table{width:100%;border-collapse:collapse;font-size:.92em}
th{background:#f0ede6;text-align:left;padding:10px 12px;border-bottom:2px solid #ddd}
td{padding:10px 12px;border-bottom:1px solid #eee;vertical-align:top;line-height:1.5}
tr.correct-row td{background:#f0f7f0}
.look-for,.lc-focus,.progression-ctx{padding:14px 18px;margin:20px 0;border-radius:6px;line-height:1.6}
.look-for{background:#fef9ec;border-left:4px solid #d4a843}
.lc-focus{background:#f1f8e9;border:1px solid #c5e1a5}
.progression-ctx{background:#f0f4ff;border:1px solid #b8c8f0}
.look-for strong,.lc-focus strong,.progression-ctx strong{display:block;margin-bottom:6px;font-size:.85em;text-transform:uppercase}
.look-for strong{color:#7a5c00} .lc-focus strong{color:#2e7d32} .progression-ctx strong{color:#1a56c4}
```

---

## Student-facing template

`<!DOCTYPE html>`, `<html lang="en">`, a head carrying `<meta charset="UTF-8">`, the viewport meta,
`<title>CFU: [Topic]</title>`, and `<style>` with the base + student rules. Then:

```html
<body>
<main>

<h1>[Topic/Concept]</h1>
<div class="name-line">Name: <span>&nbsp;</span> &nbsp;&nbsp; Date: <span>&nbsp;</span></div>

<!-- Repeat per item. Prompt and choices only — nothing else. -->
<div class="item">
  <div class="item-number">Item 1</div>
  <div class="item-prompt">
    [Prompt text — identical to the teacher file. Where the answer turns on a property only the
    figure shows, this names that property.]
  </div>
  <svg viewBox="0 0 420 110" role="img" aria-labelledby="i1-t i1-d"
       xmlns="http://www.w3.org/2000/svg">
    <title id="i1-t">[what the figure shows, in the terms the item is read in]</title>
    <desc id="i1-d">[enough detail to work the item from this text alone]</desc>
  </svg>
  <ul class="choices">
    <li><strong>A.</strong> [Choice A]</li>
    <li><strong>B.</strong> [Choice B]</li>
    <!-- C, D likewise -->
  </ul>
  <!-- Constructed response instead: a few <div class="response-line"></div> in place of the <ul> -->
</div>

</main>
</body>
</html>
```

---

## Teacher-facing template

**The prose fields are teacher notes, not assessment specs.** One thought per sentence, under about
35 words; the plain phrase leads and the technical term follows in parentheses; the student is the
subject. *"Students can locate a non-unit fraction on a number line by counting the total number of
equal parts (denominator)…"* is a teacher's note; *"Reveal whether a student can coordinate both
parts of locating…"* describes the design of the check.

Same shell, with `<title>CFU: [Topic] — Teacher Guide</title>` and the base + teacher rules.

```html
<body>
<main>

<h1>CFU: [Topic/Concept] — Teacher Guide</h1>

<div class="meta">
  <!-- One label/value span pair per row, these six in this order. Contents per references/math.md. -->
  <span class="meta-label">Standard</span>
  <span class="meta-value">[CCSS code] — [verbatim standard language from the retrieval]</span>
  <span class="meta-label">Grade</span>
  <span class="meta-value">[Grade]</span>
  <span class="meta-label">Rigor</span>
  <span class="meta-value">[Conceptual / Procedural / Application / Mixed]</span>
  <span class="meta-label">Learning goal</span>
  <span class="meta-value">[Two sentences, one thought each, under about 30 words. First: what a student who has this can do — "Students can …", scoped to the confirmed focus. Second: how that shows up in these items. Not "students understand [standard]".]</span>
  <span class="meta-label">Prerequisite standards</span>
  <span class="meta-value">[1–3 codes, each with a short gist]</span>
  <span class="meta-label">Leads to</span>
  <span class="meta-value">[1–2 codes]</span>
</div>

<div class="lc-focus">
  <strong>Focus</strong>
  [Three or four short sentences, students as the subject, one thought each: what students do in a teacher's words; the constraints the confirmed focus carries (number range, denominator set, operation, representational form) in a sentence of their own; why this focus now. A fact about the mathematics, not how it was settled.]
</div>

<div class="progression-ctx">
  <strong>Progression &amp; coherence context</strong>
  [2–3 sentences: what prerequisite understanding this assumes, what makes this standard's demand distinctive at this grade, and how that decides whether a response routes prior-grade or on-grade.]
</div>

<!-- Repeat per item -->
<div class="item">
  <!-- the same item-number / item-prompt / figure / choices markup as the student file, verbatim -->
  <h2>Correct answer &amp; interpretation guide</h2>
  <div class="correct-answer">✓ Correct answer: [Answer]</div>

  <table>
    <thead>
      <tr><th scope="col" style="width:22%">Response</th><th scope="col" style="width:40%">What it suggests</th><th scope="col">Suggested next step</th></tr>
    </thead>
    <tbody>
      <tr class="correct-row">
        <td>[Correct answer]</td>
        <td>Student [what they can now do, per the confirmed focus]</td>
        <td>[What extends or advances this]</td>
      </tr>
      <tr>
        <td>A. [Distractor]</td>
        <td>[What the student did, observably] — prerequisite gap: [prior-grade content not yet secure]</td>
        <td>[The specific prior standard code and the sub-skill to revisit]</td>
      </tr>
      <tr>
        <td>B. [Distractor]</td>
        <td>[What the student did, observably] — on-grade confusion: [which part of this demand]</td>
        <td>[Sub-skill or representation — distinct from every other row]</td>
      </tr>
      <!-- One row per remaining choice, same shape -->
    </tbody>
  </table>
</div>

<div class="look-for">
  <strong>What to look for across responses</strong>
  [2–3 sentences, grounded in the coherence map: whether a shared pattern points to one prior-grade gap those students can all be routed to, versus scattered on-grade confusions each needing different handling. Asset-based.]
</div>

</main>
</body>
</html>
```

---

Copyright 2026 Anthropic, PBC · Copyright 2026 Learning Commons · SPDX-License-Identifier: Apache-2.0
````

## File: plugin/skills/k12-check-for-understanding/references/verification.md
size: 5170
sha: e7f1eec104a6a922f2c749e7b91a687fb55b59c5
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/plugin/skills/k12-check-for-understanding/references/verification.md

```text
# Verification — k12-check-for-understanding

Read in full at Step 6, after both files are written and **before either is shown to the teacher**. Run both gates on every item; one that fails is fixed and the files rewritten before
anything is delivered. The teacher never sees this — no narration, no "I checked and found…".

**These gates run on files that already exist in `$OUTPUT_DIR`.** If you are reading this while still
composing items, you are early: finish both files, then read them back and work from that text. By
then you know which answer you meant, and a keyed item with two correct answers looks fine — so the
gates read the item back **as text**, ask you to **write things down**, and run **after** the files
exist.

If a subagent is available, hand it only the student handout — no key, no guide, no context — and
ask it to answer each item and report every choice it can defend; a fresh reader cannot rationalize
toward the answer you meant.

---

## Gate A — exactly one defensible answer

For each multiple-choice item, take every choice in turn — distractors included — and write one line
stating **the strongest case that this choice is correct**, using only what the stem and its figure
state. "Wrong because it isn't the key" is the gate failing silently. **Write the lines into a
scratch file** (`/tmp/verify_notes.md`, not `$OUTPUT_DIR`): a case you have to type out is one you
have to actually make. Then count the choices whose case survives.

- **Exactly one** → the item passes.
- **Two or more** → unanswerable as written, usually because the key rests on a property the stem
  never named. Repairs in order: (1) **name the property in the stem**, keeping the distractor — *"Which
  expression shows the number of groups times the size of each group?"* rules out the reversed factor
  order *"Which expression matches the picture?"* leaves open; (2) **remove the choice**.
- **None** → either **a figure isn't described** (if you can't tell which choice is correct from the
  stem, the choices, and the figure's `<title>`/`<desc>`, neither can a student who can't see the
  picture — give it the missing text and rerun), or **the key is wrong** and the item needs
  reworking.

Four recurring sources of a second defensible answer, worth checking directly:

- **Factor order is a convention, not a fact** — `a × b` and `b × a` are the same number, so both are
  defensible unless the stem says which factor names the group count.
- **Equivalent forms are one answer** — `1/2`, `0.5`, `2/4`, `50%` wearing four hats; two among the
  options means two correct answers.
- **A right verdict reached by invalid reasoning** — where choices are whole statements (*"No,
  because…"*), several can carry the correct verdict and differ only in justification, which works
  only when the stem asks to be judged on the reasoning.
- **Rounding, units, unstated precision** — a choice correct to the nearest whole unit is defensible
  against a key correct to the tenth unless the stem fixes the precision.

---

## Gate B — every distractor is a place a student's reasoning lands

The reverse of Gate A: does the error each distractor is meant to reveal actually produce that
choice? A distractor no described reasoning reaches is unpickable, and the student it was written for
lands elsewhere with no row waiting.

**First, compute the stem's actual quantities** — not what you intended, but the numbers as written:
first differences, ratios column by column, totals, plotted coordinates. A few lines of `python3 -c`
prints the property you were not looking for — a stem usually carries the one you chose it for and
one you didn't.

**Then, for each row of the interpretation guide**, apply the error it names to those quantities and
state which choice it produces.

- **That row's own choice** → passes.
- **A different choice** → fails. Either the choice is wrong for that error or the row's claim is;
  fix whichever is mistaken, then check the row it lands on instead.
- **The error can't be applied to this stem** → the distractor isn't reachable. Replace it with one
  this stem can produce.

The worked failure: totals 45, 70, 95, 120 have first differences 25, 25, 25 — constant. A distractor
reading *"the total paid does not increase by the same amount every time"* is false about that table,
so a student applying the additive test reads constant differences and picks a different choice.
Writing the differences down surfaces this; reading the distractor and nodding does not.

Constructed-response items have no distractors, so Gate B applies to the response types the guide
names: each is one a student could produce here.

---

## After the gates

Fix what the gates found, rewrite both files, and re-run them on anything you changed — a
repaired stem can break a distractor that was fine before. If an item still fails, **replace it
rather than patch it again**. Both files are rewritten in place in `$OUTPUT_DIR`, so the fixed
version is the first the teacher sees.

---

Copyright 2026 Anthropic, PBC · Copyright 2026 Learning Commons · SPDX-License-Identifier: Apache-2.0
```

## File: plugin/skills/k12-check-for-understanding/LICENSE
size: 11357
sha: 7a4a3ea2424c09fbe48d455aed1eaa94d9124835
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/plugin/skills/k12-check-for-understanding/LICENSE

```text

                                 Apache License
                           Version 2.0, January 2004
                        http://www.apache.org/licenses/

   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION

   1. Definitions.

      "License" shall mean the terms and conditions for use, reproduction,
      and distribution as defined by Sections 1 through 9 of this document.

      "Licensor" shall mean the copyright owner or entity authorized by
      the copyright owner that is granting the License.

      "Legal Entity" shall mean the union of the acting entity and all
      other entities that control, are controlled by, or are under common
      control with that entity. For the purposes of this definition,
      "control" means (i) the power, direct or indirect, to cause the
      direction or management of such entity, whether by contract or
      otherwise, or (ii) ownership of fifty percent (50%) or more of the
      outstanding shares, or (iii) beneficial ownership of such entity.

      "You" (or "Your") shall mean an individual or Legal Entity
      exercising permissions granted by this License.

      "Source" form shall mean the preferred form for making modifications,
      including but not limited to software source code, documentation
      source, and configuration files.

      "Object" form shall mean any form resulting from mechanical
      transformation or translation of a Source form, including but
      not limited to compiled object code, generated documentation,
      and conversions to other media types.

      "Work" shall mean the work of authorship, whether in Source or
      Object form, made available under the License, as indicated by a
      copyright notice that is included in or attached to the work
      (an example is provided in the Appendix below).

      "Derivative Works" shall mean any work, whether in Source or Object
      form, that is based on (or derived from) the Work and for which the
      editorial revisions, annotations, elaborations, or other modifications
      represent, as a whole, an original work of authorship. For the purposes
      of this License, Derivative Works shall not include works that remain
      separable from, or merely link (or bind by name) to the interfaces of,
      the Work and Derivative Works thereof.

      "Contribution" shall mean any work of authorship, including
      the original version of the Work and any modifications or additions
      to that Work or Derivative Works thereof, that is intentionally
      submitted to Licensor for inclusion in the Work by the copyright owner
      or by an individual or Legal Entity authorized to submit on behalf of
      the copyright owner. For the purposes of this definition, "submitted"
      means any form of electronic, verbal, or written communication sent
      to the Licensor or its representatives, including but not limited to
      communication on electronic mailing lists, source code control systems,
      and issue tracking systems that are managed by, or on behalf of, the
      Licensor for the purpose of discussing and improving the Work, but
      excluding communication that is conspicuously marked or otherwise
      designated in writing by the copyright owner as "Not a Contribution."

      "Contributor" shall mean Licensor and any individual or Legal Entity
      on behalf of whom a Contribution has been received by Licensor and
      subsequently incorporated within the Work.

   2. Grant of Copyright License. Subject to the terms and conditions of
      this License, each Contributor hereby grants to You a perpetual,
      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
      copyright license to reproduce, prepare Derivative Works of,
      publicly display, publicly perform, sublicense, and distribute the
      Work and such Derivative Works in Source or Object form.

   3. Grant of Patent License. Subject to the terms and conditions of
      this License, each Contributor hereby grants to You a perpetual,
      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
      (except as stated in this section) patent license to make, have made,
      use, offer to sell, sell, import, and otherwise transfer the Work,
      where such license applies only to those patent claims licensable
      by such Contributor that are necessarily infringed by their
      Contribution(s) alone or by combination of their Contribution(s)
      with the Work to which such Contribution(s) was submitted. If You
      institute patent litigation against any entity (including a
      cross-claim or counterclaim in a lawsuit) alleging that the Work
      or a Contribution incorporated within the Work constitutes direct
      or contributory patent infringement, then any patent licenses
      granted to You under this License for that Work shall terminate
      as of the date such litigation is filed.

   4. Redistribution. You may reproduce and distribute copies of the
      Work or Derivative Works thereof in any medium, with or without
      modifications, and in Source or Object form, provided that You
      meet the following conditions:

      (a) You must give any other recipients of the Work or
          Derivative Works a copy of this License; and

      (b) You must cause any modified files to carry prominent notices
          stating that You changed the files; and

      (c) You must retain, in the Source form of any Derivative Works
          that You distribute, all copyright, patent, trademark, and
          attribution notices from the Source form of the Work,
          excluding those notices that do not pertain to any part of
          the Derivative Works; and

      (d) If the Work includes a "NOTICE" text file as part of its
          distribution, then any Derivative Works that You distribute must
          include a readable copy of the attribution notices contained
          within such NOTICE file, excluding those notices that do not
          pertain to any part of the Derivative Works, in at least one
          of the following places: within a NOTICE text file distributed
          as part of the Derivative Works; within the Source form or
          documentation, if provided along with the Derivative Works; or,
          within a display generated by the Derivative Works, if and
          wherever such third-party notices normally appear. The contents
          of the NOTICE file are for informational purposes only and
          do not modify the License. You may add Your own attribution
          notices within Derivative Works that You distribute, alongside
          or as an addendum to the NOTICE text from the Work, provided
          that such additional attribution notices cannot be construed
          as modifying the License.

      You may add Your own copyright statement to Your modifications and
      may provide additional or different license terms and conditions
      for use, reproduction, or distribution of Your modifications, or
      for any such Derivative Works as a whole, provided Your use,
      reproduction, and distribution of the Work otherwise complies with
      the conditions stated in this License.

   5. Submission of Contributions. Unless You explicitly state otherwise,
      any Contribution intentionally submitted for inclusion in the Work
      by You to the Licensor shall be under the terms and conditions of
      this License, without any additional terms or conditions.
      Notwithstanding the above, nothing herein shall supersede or modify
      the terms of any separate license agreement you may have executed
      with Licensor regarding such Contributions.

   6. Trademarks. This License does not grant permission to use the trade
      names, trademarks, service marks, or product names of the Licensor,
      except as required for reasonable and customary use in describing the
      origin of the Work and reproducing the content of the NOTICE file.

   7. Disclaimer of Warranty. Unless required by applicable law or
      agreed to in writing, Licensor provides the Work (and each
      Contributor provides its Contributions) on an "AS IS" BASIS,
      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
      implied, including, without limitation, any warranties or conditions
      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
      PARTICULAR PURPOSE. You are solely responsible for determining the
      appropriateness of using or redistributing the Work and assume any
      risks associated with Your exercise of permissions under this License.

   8. Limitation of Liability. In no event and under no legal theory,
      whether in tort (including negligence), contract, or otherwise,
      unless required by applicable law (such as deliberate and grossly
      negligent acts) or agreed to in writing, shall any Contributor be
      liable to You for damages, including any direct, indirect, special,
      incidental, or consequential damages of any character arising as a
      result of this License or out of the use or inability to use the
      Work (including but not limited to damages for loss of goodwill,
      work stoppage, computer failure or malfunction, or any and all
      other commercial damages or losses), even if such Contributor
      has been advised of the possibility of such damages.

   9. Accepting Warranty or Additional Liability. While redistributing
      the Work or Derivative Works thereof, You may choose to offer,
      and charge a fee for, acceptance of support, warranty, indemnity,
      or other liability obligations and/or rights consistent with this
      License. However, in accepting such obligations, You may act only
      on Your own behalf and on Your sole responsibility, not on behalf
      of any other Contributor, and only if You agree to indemnify,
      defend, and hold each Contributor harmless for any liability
      incurred by, or claims asserted against, such Contributor by reason
      of your accepting any such warranty or additional liability.

   END OF TERMS AND CONDITIONS

   APPENDIX: How to apply the Apache License to your work.

      To apply the Apache License to your work, attach the following
      boilerplate notice, with the fields enclosed by brackets "[]"
      replaced with your own identifying information. (Don't include
      the brackets!)  The text should be enclosed in the appropriate
      comment syntax for the file format. We also recommend that a
      file or class name and description of purpose be included on the
      same "printed page" as the copyright notice for easier
      identification within third-party archives.

   Copyright [yyyy] [name of copyright owner]

   Licensed under the Apache License, Version 2.0 (the "License");
   you may not use this file except in compliance with the License.
   You may obtain a copy of the License at

       http://www.apache.org/licenses/LICENSE-2.0

   Unless required by applicable law or agreed to in writing, software
   distributed under the License is distributed on an "AS IS" BASIS,
   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
   See the License for the specific language governing permissions and
   limitations under the License.
```

## File: plugin/skills/k12-check-for-understanding/NOTICE
size: 320
sha: 4b912282e26b6f6e31990a3b86ca206ab75e015c
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/plugin/skills/k12-check-for-understanding/NOTICE

```text
Copyright 2026 Anthropic, PBC
Copyright 2026 Learning Commons

Acknowledgment
We wish to acknowledge that 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.
```

## File: plugin/skills/k12-check-for-understanding/rendering/claude-artifact.md
size: 1722
sha: f47f9953e6ade1b89f849334f2fea6e300250372
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/plugin/skills/k12-check-for-understanding/rendering/claude-artifact.md

```text
# Make the check as Claude slides

Use this when Step 5 found a **Slides** type. The check becomes a slide deck the teacher projects,
and the teacher guide becomes a Claude Doc.

`references/output.md` still sets the content: the items and their exact wording and order, what a
figure's description may say, the guide's sections, and no provenance. Its file names, styles and
HTML templates are only for the file path.

## The deck

Create it from the Slides type with the topic as its title, then follow the Slides type's own
instructions for writing and formatting it. Give each item its own slide: the prompt, any figure,
and the choices, large enough to read from the back of the room. Nothing else goes on the deck — no
answers, no notes for the teacher.

Draw a figure that has to be exact with a short Python script (matplotlib, or plain SVG), and give it
alt text by the same rule as a figure's `<desc>`: enough to work the item, never the item's work done.

## The teacher guide

Make it a Claude Doc, following Claude Docs' own instructions, with the sections of the teacher
template: each item word for word as on its slide, its answer and interpretation table, and what to
look for. Without Claude Docs functionality, write the guide as the teacher `.html` file instead.

## If it can't be made

If the deck can't be created or written, write both `.html` files as `references/output.md` says and
don't mention slides.

## Hand it over

Give the teacher the links in plain words. Don't talk about tools or formats. To revise, change the
affected slide and its place in the guide, and nothing else.

---

Copyright 2026 Anthropic, PBC · Copyright 2026 Learning Commons · SPDX-License-Identifier: Apache-2.0
```

## File: README.md
upstream: https://github.com/anthropics/k12-teacher-skills/blob/main/README.md

````text
# Agent Skills for K-12 Teachers
Source code and evaluation framework for the agent skills that are included with [Claude for Teachers](https://claude.com/solutions/teachers). 

Included are four skills:
- `k12-lesson-plan-creation`: Builds classroom-ready, standards-aligned lesson plans, optionally aligned to a teacher's curriculum
- `k12-lesson-differentiation`: Adapts an existing lesson into tiered versions (below / at / above proficiency-level) and for specific student needs, keeping core content consistent across tiers
- `k12-lesson-prep`: A prep partner for a lesson the teacher already has; works through the key student task with them and leaves a short teacher-only prep note
- `k12-check-for-understanding`: Builds a 1–3 item formative check for a math standard, with distractors drawn from documented misconceptions and a teacher guide routing each response to a next step

Where noted, materials in this repo were co-developed between Anthropic and Learning Commons, who collaborated to help Claude create classroom materials that are grounded in academic standards and follow best practices from learning science research.

The skills are written to make effective use of the Learning Commons Knowledge Graph [Claude connector](https://claude.com/connectors/learning-commons-knowledge-graph) that is included for all teachers using Claude for Teachers. This connector provides Claude with access to academic standards across all 50 states and progressions beneath them; adapt the skills if your environment does not have access to these tools.

## Quick start
There are two options to use these skills:

**In Claude for Teachers:**
If you have a Claude for Teachers account, there is no extra install needed. These skills are already installed for you.

**In other places:**
Load the plugin or skills manually. For instance, in Claude Code:

```
git clone https://github.com/anthropics/k12-teacher-skills
claude plugin marketplace add ./k12-teacher-skills
claude plugin install k12-education@k12-teacher-skills
```

## Layout

- `plugin/` - Main skill content bundled as a plugin; also includes teacher-focused 3rd party MCP servers that are available for users to enable
- `evals/` - Contains our evaluation framework and how to adapt them for your use case
````
