exam-quiz

Name: exam-quiz
Author: ZeKaiNie

258stars14forks

How to install exam-quiz?

zekainie/universal-examprep-skill/exam-quiz

$npx -y skills add ZeKaiNie/universal-examprep-skill --skill exam-quiz

Installs into the current project.

Summary

从 references/quiz_bank.json 抽取本章题目并按标准答案判分，支持选择、主观、画图、填空、判断、代码；主观题按 keywords 要点覆盖判分，连续错两次提供提示/跳过/归档。禁止现场编题。用于阶段检查或模考。

Files · 1

View on GitHub

SKILL.md

1	# exam-quiz — question drilling and grading
2
3	## Purpose
4
5	Present one chapter/phase-scoped bank item at a time, grade against its stored answer, archive wrong/skipped items through state, and return control to `exam-cram`. Never invent a question or answer.
6
7	## Activation
8
9	Use after teaching when a checkpoint is needed, or when the student asks for drills or a mock exam.
10
11	## Inputs
12
13	- Existing `references/quiz_bank.json`, whose items have `type`, answer/provenance fields, and `chapter` or `phase`; subjective items also have `keywords`.
14	- Current chapter/phase and `study_state.json` mastery/scope. An untagged item cannot enter a chapter checkpoint.
15	- Optional `difficulty` (1–5) and `difficulty_reason` from `score_difficulty.py`: a structural lower bound, never semantic truth or a per-student score.
16
17	## Workflow
18
19	1. Select only eligible bank items. Filter both `chapter` and `phase`. A missing bank is an incomplete workspace and returns to `exam-ingest`; an existing but empty usable pool produces no substitute and caps completion at `covered_unverified`.
20
21	The default source pool is mixed. Persist a student restriction and select it with `scripts/select_questions.py`; exclude and count items lacking `source_type`. Before any one-turn exception say 「⚠️ 临时覆盖你的 <scope> 范围偏好」 or `⚠️ Temporarily overriding your <scope> scope preference`; do not silently change the stored scope.
22
23	For targeted/checkpoint selection run `python "${CLAUDE_SKILL_DIR}/scripts/select_hard_questions.py" --workspace <ws> --chapter <current> -n <k>`. `--chapter` is the only exact chapter filter; `--from-chapter N` means every numeric chapter ≥N and is only for `shore_up`, never a checkpoint. Explicit cross-chapter practice may omit chapter. The selector combines structural difficulty (using `score_difficulty.py` on the fly when needed) with mistake/confusion/window mastery, mode, and stored scope. `fill_gaps` serves weak points `先易后难`, then mastered items `先难挑战`; `from_scratch` is globally `先易后难`. `shore_up` requires explicit chapter/from-chapter. Ordering is deterministic, not LLM ranking.
24
25	2. Show prompt assets first (fail-closed). For `requires_assets=true` or `maybe_requires_assets=true`, before asking, explaining, hinting, or solving, actually render every question-side `question_context` / `figure` / `diagram` / `table` asset, labelled `题面图` or `Question-side asset`. A path is not an image. Show `answer_context` / `worked_solution` only later, labelled `答案图` or `Answer-side asset`. Preserve but never display `student_attempt`: one occurrence taints the same physical path across the complete quiz, teaching, and content-unit layers, so an official-looking duplicate declaration is also unusable. Missing/unreadable files block the structured workspace; an existing asset that the UI cannot render causes an item-level skip. Prefer a safe, self-contained `full` item. `stub` and `page_reference` also require the prompt asset or original page first. Always use `python <package-root>/scripts/show_question_assets.py --workspace <ws> --id <qid> --lang <zh\|en>` so the shared three-layer policy is applied; exit 1 means skip. Do not bypass it by rendering a raw bank path yourself. See [`docs/file-format.md`](../../docs/file-format.md) §4.
26
27	3. Grade by type. `choice`: stored option. `subjective`: required `keywords`/steps with equivalent wording accepted and coverage reported. `fill_blank`: stored fill with valid synonyms. `true_false`: verdict plus one-line reason. `code`: required edits/output. `diagram`: run the standard algorithm from `render_hint`, derive the structure, then compare; teacher convention prevails.
28
29	4. Use the escape hatch. First wrong answer gets the logic gap, stored explanation, and a hint. On the second consecutive wrong answer offer view hint / skip and archive / continue.
30
31	5. Persist evidence and feedback. Before any write, if `study_state.json` is absent and Python works, run `python "${CLAUDE_SKILL_DIR}/scripts/update_progress.py" --workspace <ws> init`; only when Python truly cannot run may the generated Markdown be maintained directly. For every handled item record `record-phase-evidence --kind checkpoint --ref <qid> --outcome passed\|wrong\|skipped`; an ID alone is not mastery. Wrong/skipped items also use `python "${CLAUDE_SKILL_DIR}/scripts/update_progress.py" --workspace <ws> add-mistake --id <qid> --chapter <ch> --note <reason>`. A nonzero state command is a fail-loud write error, not permission to edit the generated view.
32
33	Before replying, pipe full verdict, gap, explanation, and source line to `python "${CLAUDE_SKILL_DIR}/scripts/notebook.py" --workspace <ws> add-entry --chapter <ch> --type feedback --id <qid> --title <gist>`. Same chapter/id replaces in place. Wrong/skipped feedback also passes `--mistake` to mirror `m

People who liked this also viewed

6 picks

Type

exam-quiz

258stars14forks

View on GitHub

How to install exam-quiz?

zekainie/universal-examprep-skill/exam-quiz

$npx -y skills add ZeKaiNie/universal-examprep-skill --skill exam-quiz

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# exam-quiz — question drilling and grading
2
3	## Purpose
4
5	Present one chapter/phase-scoped bank item at a time, grade against its stored answer, archive wrong/skipped items through state, and return control to `exam-cram`. Never invent a question or answer.
6
7	## Activation
8
9	Use after teaching when a checkpoint is needed, or when the student asks for drills or a mock exam.
10
11	## Inputs
12
13	- Existing `references/quiz_bank.json`, whose items have `type`, answer/provenance fields, and `chapter` or `phase`; subjective items also have `keywords`.
14	- Current chapter/phase and `study_state.json` mastery/scope. An untagged item cannot enter a chapter checkpoint.
15	- Optional `difficulty` (1–5) and `difficulty_reason` from `score_difficulty.py`: a structural lower bound, never semantic truth or a per-student score.
16
17	## Workflow
18
19	1. Select only eligible bank items. Filter both `chapter` and `phase`. A missing bank is an incomplete workspace and returns to `exam-ingest`; an existing but empty usable pool produces no substitute and caps completion at `covered_unverified`.
20
21	The default source pool is mixed. Persist a student restriction and select it with `scripts/select_questions.py`; exclude and count items lacking `source_type`. Before any one-turn exception say 「⚠️ 临时覆盖你的 <scope> 范围偏好」 or `⚠️ Temporarily overriding your <scope> scope preference`; do not silently change the stored scope.
22
23	For targeted/checkpoint selection run `python "${CLAUDE_SKILL_DIR}/scripts/select_hard_questions.py" --workspace <ws> --chapter <current> -n <k>`. `--chapter` is the only exact chapter filter; `--from-chapter N` means every numeric chapter ≥N and is only for `shore_up`, never a checkpoint. Explicit cross-chapter practice may omit chapter. The selector combines structural difficulty (using `score_difficulty.py` on the fly when needed) with mistake/confusion/window mastery, mode, and stored scope. `fill_gaps` serves weak points `先易后难`, then mastered items `先难挑战`; `from_scratch` is globally `先易后难`. `shore_up` requires explicit chapter/from-chapter. Ordering is deterministic, not LLM ranking.
24
25	2. Show prompt assets first (fail-closed). For `requires_assets=true` or `maybe_requires_assets=true`, before asking, explaining, hinting, or solving, actually render every question-side `question_context` / `figure` / `diagram` / `table` asset, labelled `题面图` or `Question-side asset`. A path is not an image. Show `answer_context` / `worked_solution` only later, labelled `答案图` or `Answer-side asset`. Preserve but never display `student_attempt`: one occurrence taints the same physical path across the complete quiz, teaching, and content-unit layers, so an official-looking duplicate declaration is also unusable. Missing/unreadable files block the structured workspace; an existing asset that the UI cannot render causes an item-level skip. Prefer a safe, self-contained `full` item. `stub` and `page_reference` also require the prompt asset or original page first. Always use `python <package-root>/scripts/show_question_assets.py --workspace <ws> --id <qid> --lang <zh\|en>` so the shared three-layer policy is applied; exit 1 means skip. Do not bypass it by rendering a raw bank path yourself. See [`docs/file-format.md`](../../docs/file-format.md) §4.
26
27	3. Grade by type. `choice`: stored option. `subjective`: required `keywords`/steps with equivalent wording accepted and coverage reported. `fill_blank`: stored fill with valid synonyms. `true_false`: verdict plus one-line reason. `code`: required edits/output. `diagram`: run the standard algorithm from `render_hint`, derive the structure, then compare; teacher convention prevails.
28
29	4. Use the escape hatch. First wrong answer gets the logic gap, stored explanation, and a hint. On the second consecutive wrong answer offer view hint / skip and archive / continue.
30
31	5. Persist evidence and feedback. Before any write, if `study_state.json` is absent and Python works, run `python "${CLAUDE_SKILL_DIR}/scripts/update_progress.py" --workspace <ws> init`; only when Python truly cannot run may the generated Markdown be maintained directly. For every handled item record `record-phase-evidence --kind checkpoint --ref <qid> --outcome passed\|wrong\|skipped`; an ID alone is not mastery. Wrong/skipped items also use `python "${CLAUDE_SKILL_DIR}/scripts/update_progress.py" --workspace <ws> add-mistake --id <qid> --chapter <ch> --note <reason>`. A nonzero state command is a fail-loud write error, not permission to edit the generated view.
32
33	Before replying, pipe full verdict, gap, explanation, and source line to `python "${CLAUDE_SKILL_DIR}/scripts/notebook.py" --workspace <ws> add-entry --chapter <ch> --type feedback --id <qid> --title <gist>`. Same chapter/id replaces in place. Wrong/skipped feedback also passes `--mistake` to mirror `m

People who liked this also viewed

6 picks

Type