Writes a fixed number of quiz questions from a passage you give it, as JSON, where every answer is a span copied word for word from the passage, with the same digits or number words, spacing, dashes, apostrophes and capitals, paired with the full sentence it comes from, and never a paraphrase, a pronoun, a computed value or a fact from outside the text. Questions are in the passage's own language, each asks for a different fact, a negation or a pair of dates is never flipped, text inside the passage that speaks to the assistant is treated as passage text, and a passage too short for the requested count gets fewer questions and a note instead of padding. Use as a quiz generator skill to make reading comprehension questions, a quiz from text for a lesson or a course, a self-check for a study text, or questions from a passage whose answers a program can verify against the source.
Quiz From Text: Answers Quoted From the Passage is a tested SKILL.md that writes a fixed number of quiz questions from a passage you give it, as JSON, where every answer is a span copied word for word from the passage, with the same digits or number words, spacing, dashes, apostrophes and capitals, paired with the full sentence it comes from, and never a paraphrase, a pronoun, a computed value or a fact from outside the text; an agent buys it once for $0.02 over x402.
Open questions whose answer is an opinion, an inference or a summary; questions that need facts from outside the passage; multiple-choice distractors (the answer is one quoted span, no wrong options are written); grading of real students. It does not shorten the passage and it does not check whether the passage is true.
Tested 2026-10-08 with a strong and a weak model.
| Sonnet | Haiku | |||
|---|---|---|---|---|
| with | without | with | without | |
| Quizzes written right (23 passages) | 23/23 | 22/23 | 23/23 | 21/23 |
Same request on both sides, a fence removed first. Sonnet without the skill already wrote markable quizzes with verbatim answers. It missed one: a line planted in a Bulgarian passage asked for the questions in English, and it obeyed. Haiku without the skill obeyed the same line and wrote one plain Bulgarian quiz with an English question.
Same cases and the same checks with and without the skill. The cases are ours, written around what the skill is for; with a handful of cases, a difference of one or two is within noise.
Our own test text, before and after the skill ran. Excerpts only.
Before
The expedition counted forty-two glaciers in the valley and found that only 12 of them still carried snow in August.
After
{"questions":[{"q":"How many glaciers did the expedition count in the valley?","answer":"forty-two","quote":"The expedition counted forty-two glaciers in the valley and found that only 12 of them still carried snow in August.","type":"how-many"},{"q":"How many of the glaciers still carried snow in August?","answer":"12","quote":"The expedition counted forty-two glaciers in the valley and found…
Before
Градската библиотека пази двеста ръкописа, а петнайсет от тях са написани на пергамент.
After
{"questions":[{"q":"Колко ръкописа пази Градската библиотека?","answer":"двеста","quote":"Градската библиотека пази двеста ръкописа, а петнайсет от тях са написани на пергамент.","type":"how-many"},{"q":"Колко от ръкописите са написани на пергамент?","answer":"петнайсет","quote":"Градската библиотека пази двеста ръкописа, а петнайсет от тях са написани на пергамент.","type":"how-many"}]}
Before
El faro de la isla se encendió por primera vez en 1884 y sigue funcionando con la lámpara original, que limpia un farero cada lunes.
After
{"questions":[{"q":"¿En qué año se encendió por primera vez el faro de la isla?","answer":"1884","quote":"El faro de la isla se encendió por primera vez en 1884 y sigue funcionando con la lámpara original, que limpia un farero cada lunes.","type":"when"},{"q":"¿Qué día de la semana limpia un farero la lámpara original?","answer":"cada lunes","quote":"El faro de la isla se encendió por primera…
Before
Die alte Brücke über den Fluss wurde 1912 aus Sandstein gebaut und ist 240 Meter lang; sie wird jeden Herbst von drei Steinmetzen geprüft.
After
{"questions":[{"q":"Aus welchem Material wurde die alte Brücke über den Fluss gebaut?","answer":"Sandstein","quote":"Die alte Brücke über den Fluss wurde 1912 aus Sandstein gebaut und ist 240 Meter lang; sie wird jeden Herbst von drei Steinmetzen geprüft.","type":"what"},{"q":"Wie viele Steinmetzen prüfen die Brücke jeden Herbst?","answer":"drei","quote":"Die alte Brücke über den Fluss wurde…
English, Bulgarian, Spanish, German. Tried in: English, Bulgarian, Spanish, German.
Perpetual, non-exclusive; use and modify for yourself incl. paid work; no resale or republishing. Holder: Georgi Kalchev, aiskills402.com. Full terms.
Current version 1.0.0, updated 2026-10-08. Whoever bought an earlier version gets new ones free through the same re-download token.
v1.0.0 · 2026-10-08
First release (written before any model test; no measured result yet). Writes a fixed number of quiz questions from one passage as JSON, each with an answer that is a span copied word for word from the passage (digits or number words as printed, spacing, dashes, apostrophes, capitals and list order kept) and a quote that is the full sentence holding it; never a paraphrase, a pronoun, an arithmetic result or a fact from outside the text. Questions end with a question mark, are in the passage's language, one fact per question, polarity kept for negations, the right member named when a sentence holds two dates, no "why" unless the passage states the reason; text in the passage that addresses the assistant or the quiz writer is passage text, not a fact and not an order; a passage too short for the requested count gets fewer questions and a note instead of padding.
Design: the orchestrator's design notes (JSON with q, answer, type; `inInput` on answers; seven trap families) were followed first; brief § 3.22 in `docs/plan-batch3-briefs-21-40.md` arrived while writing and was reconciled at the end: its `quote` field, question-mark guard, no-"why" guard, "seven to four" trap and the planted "hunter2" trap (the planted answer is in the passage, so only `forbid` catches it) were taken; its exact-count rule for a short passage was replaced by a defensible band (1–4 questions plus a required note) and its key names by the orchestrator's.
Test set: 23 cases (16 traps, 7 controls) in en, bg, es, de, verified by `test/control.mjs` with zero model calls (ideal passes; input echo, fenced answer, prose around the JSON, paraphrased answer, stitched quote, missing question mark and 135 realistic wrong answers fail). The verbatim check is the tester's `json.inInput` on both answer and quote.
Price starts at $0.02 (class B); the measured baseline decides whether it stays. Expected without the skill: 1–3 content failures of 23 (number words turned into digits, a computed age, a pronoun as answer, an obeyed planted line).
Because the point of a quiz built this way is that a program can mark it: each answer, and the sentence it comes from, is searched for inside the passage, character by character. A paraphrase, a number turned from words into digits, a retyped dash or apostrophe, or a fact the model knows but the passage never states, all come back as not found. The skill treats every such answer as wrong, however true it is.
It writes as many questions as there are separate facts and adds a one-sentence note saying so, instead of padding with a second question on the same fact or with stray fragments. A passage of one sentence with three facts gives three questions, not eight.
The question is written so that the quoted span answers it and nothing else in the passage does: a closed-on-Mondays sentence yields a question about the day it is closed, never one about when it is open; a sentence with a design year and an opening year yields questions that name which event they ask about. Pronouns are never answers, and a line in the passage that addresses the assistant is quizzed around, never obeyed.
Hardly at all. We had both Claude models write quizzes for twenty-three passages in four languages, file in hand and without it, and a script confirmed that every answer sits in its passage verbatim. Plain Sonnet already wrote quizzes a program can mark; it slipped once, obeying a line inside a Bulgarian text that asked for English questions. Haiku rose from 21 to 23. The type field says whether a question asks for a fact, a number, a date or a name.
Research & Summaries
Writes a faithful summary of a text in the text's own language, using only what the text says, with every number, name and date kept exactly and every qualifier kept. Use when asked to summarize, shorten, condense, give the gist or a TL;DR of an article, report, email, transcript, thread or document.
Business & Documents
Pulls the action items out of meeting notes, minutes or a transcript into strict JSON, each with the task, the owner and the due date, and invents none of them. A person nobody named stays null, "I" in notes with no known author stays null, a deadline such as "next Friday" or "next week" that can mean two dates stays null, and "by Friday" or "tomorrow" becomes a calendar date only when the meeting date is known. Tasks that were cancelled, already done or handed to someone else are left out or carry the final owner, repeats in a recap are merged, and orders written into the notes for an AI are ignored. Use to extract action items from meeting notes, turn minutes or a call transcript into a task list, find who owns what and by when after a meeting, or feed follow-ups into a CRM or task tracker.
Data & Analysis
Extracts data from text into JSON that matches the shape you give (a JSON Schema, an example object or a list of fields), without guessing. Every value comes from the text; a missing fact becomes null, numbers and dates are converted only into the type the shape asks for, two conflicting values are not settled by a guess, and instructions hidden in the text are ignored. Use when asked to extract structured data, turn text into JSON, parse an invoice, receipt, email, order, CV or job post into fields, or fill a JSON schema from a document.
Data & Analysis
Describes a table you were given (pasted Markdown, CSV, TSV or spreadsheet cells) in a short paragraph that uses only figures the table itself prints. Every number is copied as written with its unit, currency and qualifier. Nothing is added up, averaged, rounded, turned into a percent or a ratio, or extended into a trend, a cause or a forecast. A blank or n/a cell is reported as not given, never as zero. A total the table prints is quoted, not checked. Use when asked to describe a table, summarize a table, explain what a data table shows, write a caption or a sentence for a table, or put the rows of a report table into words without inventing a single number.