Mom Frq

{
  "source": "ch6-questions.txt",
  "questions": {
    "1": { "status": "completed", "file": "q1-single-proportion-hypothesis-test.php", "title": "...", "prompt": "..." },
    "3": { "status": "pending", "file": null, "title": "...", "prompt": "..." }
  }
}

Rules:

• Read the manifest before doing anything in both single and batch mode.
• Never overwrite completed questions. If status is "completed", skip it entirely.
• After writing a question file, update the manifest: set status to "completed" and file to the filename (just the filename, not the full path).
• When the user says "all" or gives a range, filter to only "pending" entries.
• Output folder: Derive the subfolder from the manifest's source field — strip the .txt extension to get the folder name (e.g., ch6-questions.txt → ch6-questions, butte-stats-week1-5-questions.txt → butte-stats-week1-5-questions). All question files for this manifest go into C:/Users/shuff/mom/questions/frq/<source-folder>/. Create the folder if it doesn't exist.

Batch Mode (skip rubric review)

When the user provides multiple questions and indicates they want to skip rubric review (e.g., "batch", "all", "skip review", "no review", or passes a range like "questions 3-8"), use this workflow:

1. Read the manifest (C:/Users/shuff/mom/questions/manifest.json) to identify which questions are pending.
2. Check system resources to determine max parallel agents. Run:

   nproc && cat /proc/meminfo | head -2
   

   Then calculate: max_parallel = min(cores / 8, available_ram_gb / 4, 4), floored at 2, capped at 4. If the check fails, default to 3. If there are more pending questions than max_parallel, process them in waves of max_parallel at a time (launch wave 1, wait for all to finish and update manifest, then launch wave 2, etc.).
3. Read one existing completed question file to use as a reference for sub-agents.
4. Spawn agents up to max_parallel at a time using the Task tool with subagent_type: "general-purpose". Each sub-agent gets:
   • The full SKILL.md content (this file) as context for the template pattern and all rules
   • The specific question title and prompt from the manifest
   • The contents of the reference question file
   • Instructions to design the rubric AND generate the code in one shot (no user review)
   • The output file path: C:/Users/shuff/mom/questions/frq/<source-folder>/q<number>-<kebab-case-topic>.php where <source-folder> is the manifest's source field with .txt stripped (e.g., ch6-questions). So a full example path: C:/Users/shuff/mom/questions/frq/ch6-questions/q3-chi-square-goodness-of-fit.php. Create the folder if it doesn't exist.
5. After each agent finishes, update the manifest entry: set status to "completed" and file to the filename written.
6. If more pending questions remain, launch the next wave (back to step 4).
7. After all waves finish, report a summary table:

| # | Question | File | Status |
|---|----------|------|--------|
| 3 | Chi-Square Goodness of Fit | q3-chi-square-goodness-of-fit.php | Done |
| 4 | Chi-Square Independence | q4-chi-square-independence.php | Done |
| 1 | Single Proportion Inference | q1-single-proportion-hypothesis-test.php | Skipped (already completed) |

Sub-agent prompt must include:

• The entire Required Template Pattern, all syntax rules, rubric design rules, and common mistakes from this skill
• A reference question file (read its contents and include them)
• The instruction: "Design the rubric internally (2-4 categories, 10 points total, neutral student items, specific ideal targets) and generate the complete MOM code. Write the file directly. No user interaction needed."
• All style rules: write like a human, no em dashes, natural tone

When to use batch mode vs. single mode:

• User gives ONE question → use the Two-Step Workflow below (rubric review)
• User gives MULTIPLE questions or says "skip review" / "batch" / "no review" → use Batch Mode above

Two-Step Workflow (single question, default)

NEVER skip to code generation. Always do Step 1 first and wait for approval.

Step 1: Generate Rubric for Approval

Given a topic/description, generate a rubric proposal and present it to the user BEFORE writing any MOM code. Use this exact format:

**Essay Prompt:** [The question students will answer]

**Randomized Contexts:** [3 scenario variations that will rotate]

**Rubric Categories:**

| Category | Student Checklist Items | Ideal Answer Targets | Points |
|----------|----------------------|---------------------|--------|
| [Name]   | - Checkbox item 1    | Target: "..."       | N      |
|          | - Checkbox item 2    | Target: "..."       |        |
| [Name]   | - Checkbox item 1    | Target: "..."       | N      |
|          | - Checkbox item 2    | Target: "..."       |        |
| [Name]   | - Checkbox item 1    | Target: "..."       | N      |

**Total:** N points

**Model Narrative Response:**
[Full model answer that a perfect student response would look like]

Rubric generation rules:

1. Use as few or as many categories as the question needs. Each category represents one conceptual component of the answer. A simple "explain this concept" question might only need 2 categories; a multi-part analysis could need 4. The audience is high school seniors / intro stats students, so keep categories focused and don't over-slice simple ideas.
2. Each category has 1-3 checklist items. Vary the count based on how much each category covers. A straightforward category might need just 1 item; a meatier one might need 3. Don't default to 2 items per category out of habit.
3. Student items are NEUTRAL. They describe what to address, NOT the correct answer. Example: "Describe the sampling method" not "Take many random samples."
4. Ideal targets are SPECIFIC. They state the exact correct answer for grading. Example: Target: "Take many random samples and calculate the mean of each."
5. Points per category must sum to the total. Default total is 10 unless user specifies otherwise.
6. Model narrative is ONE paragraph (2-5 sentences) that covers all categories. Uses <b>bold</b> for key phrases that map to rubric targets.
7. Randomized contexts should produce meaningfully different scenarios while keeping the same rubric structure.
8. Write like a human. All generated text (rubric items, ideal targets, model narratives, question prompts) should sound natural and conversational, like a good instructor wrote it. Avoid stiff, overly formal, or robotic phrasing.
9. No em dashes. Never use —, —, or the literal em dash character anywhere in generated code or text. Use commas, periods, semicolons, or just restructure the sentence instead.

After presenting the rubric, ASK: "Does this rubric look correct? Any categories, items, or point values to adjust?"

Wait for user approval before proceeding to Step 2.

Step 2: Generate MOM Code

Only after the user approves (or adjusts) the rubric, generate the complete MOM question code following the Required Template Pattern below.

Write the code to a single file in C:/Users/shuff/mom/questions/frq/<source-folder>/ using the format q<number>-<kebab-case-topic>.php (e.g., q1-single-proportion-hypothesis-test.php), where <source-folder> is the manifest's source field with .txt stripped. Create the folder if it doesn't exist. The file should contain all sections in order, clearly commented:

// === NAME - DESCRIPTION: [Short title] - [1-2 sentence summary] ===
// === SET QUESTION TYPE TO: multipart ===

// === COMMON CONTROL (paste into Common Control) ===
[all Common Control code here]

//question text

$questiontext
$answerbox[0]

///

$rubricanswerbutton

After writing the file:

1. Update the manifest (C:/Users/shuff/mom/questions/manifest.json): set the question's status to "completed" and file to the filename.
2. Tell the user the file path, to set the question type to multipart, and the suggested name/description (format: Name - Description).

Map the approved rubric directly:

• Each category → one <tr> row in both rubric tables
• Each checklist item → one <li><label><input type="checkbox"> in student rubric
• Each ideal target → one <span class="ideal-ans">Target: "..."</span> in instructor rubric
• Narrative variables → one $r_ variable per category
• Model narrative → $sample_narrative composed from $r_ variables

Question Structure

Every free response question has these sections in the MOM editor. Set the question type to multipart. Use /// to separate Question Text from Answer in a single block:

Common Control  →  All code: variables, CSS, rubrics, question text variable
Question Text   →  $questiontext and $answerbox[0], then /// delimiter, then $rubricanswerbutton

Put EVERYTHING in Common Control. Question Text/Answer only references variables, separated by ///.

Required Template Pattern

Follow this exact structure. Numbered sections in Common Control:

//common control (mandatory)

loadlibrary("stats");

$anstypes = array("essay");
$displayformat[0]='editornopaste';

/* ---------- 1. Dynamic Context Generation ---------- */
// Randomized topic/context variables
$contexts = array("context A", "context B", "context C");
$i = rand(0, count($contexts)-1);
$topic = $contexts[$i];

// Narrative variables for the model answer
$r_point1 = "explanation of first key concept";
$r_point2 = "explanation of second key concept";
$r_point3 = "explanation of the outcome or conclusion";

$sample_narrative = "Model narrative using <b>$r_point1</b> and connecting to $topic. Then <b>$r_point2</b>. Finally, <b>$r_point3</b>.";

/* ---------- 2. SHARED CSS & JS ---------- */
$css_block = '
<style>
    .rubric-container { width:100%; font-family:Arial; font-size:medium; margin:1em 0; }
    .rubric-container details { width:100%; border:1px solid #ccc; border-radius:8px; overflow:hidden; background:#fff; }
    .rubric-container summary { cursor:pointer; display:block; width:100%; background:#f8f8f8; color:#333; padding:0.35em 0.6em; font-weight:bold; border-bottom:1px solid #ccc; list-style:none; border:none; }
    .rubric-container details[open] summary { box-shadow: inset 0 -1px 0 #ccc; }
    .rubric-container summary::-webkit-details-marker { display:none; }
    .arrow-open { display:none; }
    .rubric-container details[open] .arrow-closed { display:none; }
    .rubric-container details[open] .arrow-open { display:inline; }
    .rubric-content { overflow:hidden; max-height:0; opacity:0; transition:max-height 300ms ease-out, opacity 300ms ease-out, padding 200ms ease-out; margin-top:0; background:#fafafa; box-sizing:border-box; padding:0 0.75em; }
    .rubric-container details[open] .rubric-content { max-height:2000px; opacity:1; padding:0.75em; }
    .rubric-table { border-collapse:separate; border-spacing:0; width:100%; border:1px solid #ccc; border-radius:8px; overflow:hidden; font-family:Arial; font-size:small; margin-top:10px; }
    .rubric-table th { background:#f2f2f2; padding:8px; text-align:center; }
    .rubric-table td { padding:10px; border-bottom:1px solid #eee; vertical-align:top; }
    .row-colored { background:#fff9ea; }
    .col-header { width:25%; border-top-left-radius:8px; }
    .col-check { border-top-right-radius:8px; }
    .col-cat-bot { border-bottom-left-radius:8px; }
    .col-check-bot { border-bottom-right-radius:8px; }
    .ideal-ans { display:block; background-color:#e8f5e9; font-style:italic; font-weight:bold; font-size:0.95em; margin:5px 0 10px 0; border-left:3px solid #4CAF50; padding-left:8px; }
    .full-response-box { margin-top:15px; border:2px solid #4CAF50; background-color:#e8f5e9; padding:15px; border-radius:5px; }
</style>
<script>
document.addEventListener("DOMContentLoaded", function() {
  var details = document.querySelectorAll(".rubric-container details");
  details.forEach(function(det) {
    var content = det.querySelector(".rubric-content");
    det.addEventListener("toggle", function() {
      if (det.open) {
        content.style.maxHeight = content.scrollHeight + "px";
        content.style.opacity = "1";
      } else {
        content.style.maxHeight = content.scrollHeight + "px";
        content.offsetHeight;
        content.style.maxHeight = "0";
        content.style.opacity = "0";
      }
    });
    content.addEventListener("transitionend", function() {
      if (!det.open) content.style.maxHeight = null;
    });
  });
});
</script>';

/* ---------- 3. Student Rubric (Neutral Checklist) ---------- */
// NO answers shown - just checkbox criteria for students
$rubricbutton = $css_block . '
<div class="rubric-container">
  <details>
    <summary>
      <span class="arrow-closed">&#9656;</span><span class="arrow-open">&#9662;</span>
      Click to View Grading Checklist
    </summary>
    <div class="rubric-content">
      <p style="margin:0 0 0.5em 0;"><b>Grading Criteria</b> -- ensure your explanation covers these points:</p>
      <table class="rubric-table">
        <tbody>
          <tr>
            <th class="col-header">Category</th>
            <th class="col-check">Requirement</th>
          </tr>
          <!-- ROWS: One per rubric category with checkbox items -->
          <tr class="row-colored">
            <td style="text-align:center;"><b>Category Name</b></td>
            <td>
              <ul style="list-style:none; margin:0; padding-left:0;">
                <li><label><input type="checkbox"> Requirement description.</label></li>
              </ul>
            </td>
          </tr>
          <!-- Last row uses col-cat-bot and col-check-bot classes -->
        </tbody>
      </table>
    </div>
  </details>
</div>';

/* ---------- 4. Instructor Rubric (With Answer Targets) ---------- */
// Shows ideal answers and full model response - only visible when grading
$rubricanswerbutton = $css_block . '
<div class="rubric-container">
  <details>
    <summary>
      <span class="arrow-closed">&#9656;</span><span class="arrow-open">&#9662;</span>
      Rubric &amp; Model Response
    </summary>
    <div class="rubric-content">
      <table class="rubric-table">
        <tbody>
          <tr>
            <th class="col-header">Category</th>
            <th class="col-check">Checklist &amp; Ideal Targets</th>
          </tr>
          <!-- ROWS: Same categories but with ideal-ans spans -->
          <tr class="row-colored">
            <td style="text-align:center;"><b>Category Name</b></td>
            <td>
              <ul style="list-style:none; margin:0; padding-left:0;">
                <li>Criterion label.
                    <span class="ideal-ans">Target: "The ideal answer text."</span></li>
              </ul>
            </td>
          </tr>
        </tbody>
      </table>
      <div class="full-response-box">
        <span style="color:#2E7D32; font-weight:bold;">Model Narrative Response:</span><br><br>
        '.$sample_narrative.'
      </div>
    </div>
  </details>
</div>';

/* ---------- 5. Question Text ---------- */
$questiontext = '
<div style="font-family:Arial; font-size:medium; line-height:1.6;">
  <p>Context paragraph using '.$topic.' with scenario setup.</p>
  <p><b>Essay Prompt:</b><br>
  The main question the student must answer.</p>
  <p>In your explanation, be sure to cover:</p>
  <ul>
    <li>First required point.</li>
    <li>Second required point.</li>
    <li>Third required point.</li>
  </ul>
  '.$rubricbutton.'
</div>';

Question Text and Answer (single code block separated by ///)

Important: Set the question type to multipart in the MOM editor. The /// delimiter separates Question Text from the Answer section.

$questiontext
$answerbox[0]
///
$rubricanswerbutton

MyOpenMath Syntax Rules

These are NOT standard PHP. Violating these produces broken questions.

Randomization Patterns

// Pick from a list of contexts
$contexts = array("context A", "context B", "context C");
$i = rand(0, 2);
$topic = $contexts[$i];

// Random numbers
$n = rand(25, 50);
$mean = rrand(60, 80, 0.1);

// Random names
$name = randname();

// Multiple different values
$a, $b = nonzerodiffrands(-8, 8, 2);

Rubric Design Rules

1. In single mode, rubric is generated in Step 1 and approved before coding. In batch mode, sub-agents design the rubric internally and proceed directly to code.
2. Student rubric ($rubricbutton): Neutral checklist with <input type="checkbox">. NO answers. Maps 1:1 from approved Step 1 checklist items.
3. Instructor rubric ($rubricanswerbutton): Same categories + <span class="ideal-ans">Target: "..."</span> with ideal answers. Maps 1:1 from approved Step 1 ideal targets. Includes $sample_narrative in a .full-response-box.
4. Both use the SAME $css_block (prepended to each).
5. Both use the <details>/<summary> collapsible pattern.
6. Student rubric button text: "Click to View Grading Checklist"
7. Instructor rubric button text: "Rubric & Model Response"
8. Categories, items, and targets in the code MUST match the approved rubric exactly. No adding, removing, or rewording during code generation.

Narrative Variable Pattern

Break the model answer into named concept variables, then compose:

$r_process = "what the student should describe about the process";
$r_mechanism = "what explains why/how it works";
$r_result = "what the final outcome or conclusion should be";

$sample_narrative = "Full model response using <b>$r_process</b>. Then <b>$r_mechanism</b>. Finally, <b>$r_result</b>.";

This makes the model answer:

• Easy to maintain (change one variable, narrative updates)
• Consistent between the narrative and the rubric targets
• Adaptable to randomized contexts (reference $topic in the narrative)

Strict Rules

1. No custom macros. You may ONLY use built-in MyOpenMath functions. Do NOT invent functions. If unsure whether a function exists, check the reference files first.
2. Follow official question writing syntax. Read the reference files in C:/Users/shuff/mom/ for the authoritative list of available functions:
   • myopenmath-question-writing-reference.json — full macro/type reference
   • myopenmath-libs.json — available libraries
   • help_writing.html — official help documentation
   • free-reponse-template.php — canonical template to match
3. All HTML/CSS/JS is allowed in string variables — MyOpenMath renders HTML in question text. But the code sections only use MOM syntax.
4. When in doubt, read the reference files. If you are unsure about a function signature, parameter, or whether something exists — look it up. Do not guess.

Common Mistakes

Quick Reference: Question Text & Answer Layout

Set question type to multipart. Use /// to separate Question Text from Answer:

//question text

$questiontext
$answerbox[0]

///

$rubricanswerbutton

That's it. Everything else is built in Common Control.