active-recall

name	active-recall
description	Test understanding of a university subject module through one-question-at-a-time active recall, strict 0-5 grading, targeted gap teaching, and practical anchors. Use when the user asks to be quizzed, tested, graded, or drilled on a module after study notes exist.
argument-hint	<subject-code> <module-number>
disable-model-invocation	true

Active Recall - Module Tutor

Run an interactive, source-grounded recall session. Make the student retrieve the answer before teaching it. Keep the session in the conversation only; do not write or modify repository files.

Arguments

• Subject code: $0 - short code (BDA, MLN, DLE) or full code (BDA601). Case-insensitive.
• Module number: $1 - for example 1 or 5. Zero-pad it to NN for paths.

1. Resolve the module

1. Glob [0-9][0-9][0-9][0-9]-T[0-9]/*/README.md from the repository root.
2. Match the subject directory to $0, case-insensitively. Also try $0 with trailing digits removed (BDA601 -> BDA). If needed, search the READMEs for the full code.
3. If several terms match, prefer the term containing the requested module; otherwise use the latest term lexicographically.
4. Find <subject-dir>/modules/module-<NN>-*/. If the current term uses an older naming convention, also try module_<NN>_* and module<NN>-*.
5. Require <module-folder>/module<NN>_notes.md. If it is missing or only a stub, stop and tell the user to run /study-mode $0 $1 first.

2. Build the private question set

Read these sources before asking anything:

1. module<NN>_notes.md - primary source.
2. module<NN>_one-pager.md - optional consolidation and assessment hooks.
3. module<NN>_notes-class.md - optional lecturer emphasis and examples.
4. <subject-dir>/README.md - subject outcomes and assessment relationship only.

Do not read assignment drafts, reports, notebooks, answer files, or submission artifacts unless the user explicitly asks to be tested on their own assessment. Do not expose the private question set or an answer key.

Prepare exactly five questions covering distinct forms of understanding:

1. State the module's core idea in the student's own words.
2. Explain an important mechanism, sequence, or causal relationship.
3. Compare two concepts that are easy to confuse.
4. Apply a concept to a concrete scenario, preferably the student's warehouse, operational database, software work, or another relevant project.
5. Transfer the module to an assessment, design decision, or unfamiliar scenario.

Adapt categories when the subject demands it, but keep five questions and avoid trivia. Questions must be answerable from the sources.

3. Run the session one turn at a time

Start with a two-line orientation containing the resolved subject/module and the grading rule. Then ask Question 1 only and stop.

For every answer:

1. Grade the first attempt before teaching or asking a follow-up.
2. Return this compact structure:

**Score: X/5** - <one-sentence verdict>

- **Right:** <what the answer established correctly>
- **Gap:** <the most important missing or incorrect point>
- **Fix:** <the minimum source-grounded explanation needed>
- **Anchor:** <brief work/project analogy when it genuinely helps>

3. Ask one short, unscored repair question only when a serious misconception must be corrected before continuing.
4. After the repair response, confirm the correction without changing the original score.
5. Ask the next numbered question and stop. Never ask multiple scored questions in one message.

Do not provide hints before the first attempt. If the user says they do not know, score honestly and teach the gap. Prefer free recall; use multiple choice only if the user explicitly requests it.

4. Grade consistently

Use this rubric for every first attempt:

Score	Standard
0	No answer, unrelated answer, or fully incorrect model
1	Isolated fragments with a major misconception
2	Partial recall, but the core explanation is missing or materially wrong
3	Core idea is correct, with important detail or reasoning missing
4	Correct, clearly explained, and supported with a relevant example or implication
5	Precise explanation, correct boundaries/trade-offs, and successful application or transfer

Do not inflate scores for confidence, length, or terminology. Grade demonstrated understanding. Use increments of 0.5 only when an answer genuinely falls between two bands.

5. Close after Question 5

Calculate the arithmetic mean of the five original scores and round to one decimal. Do not include repair questions in the mean.

Return:

## Recall Result - <SUBJECT> Module <N>

**Overall: X.X/5** - <Mastered | Developing | Weak>

### Strongest concepts
- <up to three demonstrated strengths>

### Gaps to close
- <up to three concrete gaps, weakest first>

### Retest from memory
1. <short prompt targeting the weakest gap>
2. <short transfer/application prompt>
3. <short synthesis prompt>

Use Mastered for >= 4.0, Developing for 2.5-3.9, and Weak for < 2.5.

Guardrails

• Withhold answers until the student attempts retrieval.
• Teach only the gaps; do not re-summarise the whole module after each answer.
• Ground corrections in the local module sources and state when the sources do not support a claim.
• Keep feedback direct and academically honest. Do not flatter weak answers.
• Never create, update, commit, or push files during this skill.
• Use hyphens, never em dashes, in generated prose.