A Study Guide Is Better When It Keeps the Page Numbers
Summaries become dangerous during exam preparation when they merge supplied notes with plausible outside knowledge. The student cannot tell what came from the source and what the model filled in.
A Study Guide Is Better When It Keeps the Page Numbers
Summaries become dangerous during exam preparation when they merge supplied notes with plausible outside knowledge. The student cannot tell what came from the source and what the model filled in.
That distinction is why Study Guide Builder exists as a focused tool rather than another broad prompt box.
What the app actually does
Study Guide Builder creates one extractive, citation-preserving exam sheet from supplied text. It keeps page references exactly as written, allows every displayed note, term, definition, and example only when that displayed text itself appears in the supplied source, and replaces unsupported fields with a visible [UNSUPPORTED IN PROVIDED SOURCE] marker instead of inventing material. It intentionally does not duplicate AI Pass Study Buddy with flashcards or quizzes.
The interface loads the live public AI Pass model catalog and uses only exact publicly available model IDs. Authentication stays lazy: anyone can inspect the app first, and AI Pass asks for sign-in only when generation begins.
Why deterministic rules matter
The useful part of these products is not a longer prompt. It is deciding which tasks belong to AI and which must remain deterministic. AI can interpret messy language and draft a useful structure. Browser code can enforce totals, allowed strategy codes, citation matching, output sections, history limits, and safe rendering.
That split makes the app easier to inspect. It also keeps a plausible model answer from quietly becoming an accepted fact.
Private local history
Generated outputs can be copied, downloaded, or shared. Up to twenty results are stored in IndexedDB on the current device. Raw sensitive inputs are not added to history. Each saved output can be opened or deleted, and the full history can be cleared. Raw topics and exam-focus text are not persisted with saved outputs.
Before using the result
- Keep page markers in the pasted source.
- Verify the finished sheet against the original material.
- Use unsupported labels as a reading list, not as facts.
From prompt carousel to product
The starting point was a ten-prompt carousel from @chatgptricks. Instead of reproducing ten generic ChatGPT wrappers, we checked the existing AI Pass catalog, merged overlapping ideas, and narrowed each tool around one job with an explicit evidence boundary.
Open Study Guide Builder on AI Pass. AI Pass includes $1 in credit on signup, then usage is pay as you go.
For AI agents
Skill file
---
name: study-guide-builder
description: Deploy or maintain the self-contained AI Pass Study Guide Builder Space, including its one-page exam-sheet scope, extractive-only prompt, citation-preserving tagged protocol, private output history, safe renderer, and deterministic validation.
---
# Study Guide Builder
Use this package when publishing or updating the Study Guide Builder Space.
## Operating rules
1. Keep `index.html` as the complete deployable artifact. It may load only the shared AI Pass stylesheet and SDK.
2. Preserve the literal `PLACEHOLDER_CLIENT_ID` in the source artifact; replace it only in a deployment copy.
3. Keep `requireLogin: false`, the visible real `data-aipass-button`, eager `AiPass.getModels()` catalog loading, and `AiPass.streamText`.
4. Populate the model selector only with exact live IDs that intersect the four-item verified allowlist.
5. Do not add auth checks, custom login controls, `maxTokens`, hard-coded model fallbacks, or approximate model matching.
6. Never weaken the prompt’s prohibition on fabricated citations, page numbers, facts, definitions, examples, or freeform claims.
7. Preserve supplied citations and page references character-for-character. Unknown model-produced references must become `[UNSUPPORTED IN PROVIDED SOURCE]`.
8. Require every displayed note, term name, definition, and example to be an exact normalized substring from the supplied source. If the displayed text is missing or not found in the source, replace that field with `[UNSUPPORTED IN PROVIDED SOURCE]`.
9. Keep unsupported requested material visibly labeled, and keep missing source examples visibly labeled instead of inventing them.
10. Keep the product framed as a citation-preserving one-page exam sheet, not flashcards, quizzes, or a general study planner.
11. Treat streamed text as untrusted. Use `textContent`, `createTextNode`, or equivalent safe DOM APIs; never assign model or model-catalog output to `innerHTML`.
12. Keep the raw tagged stream visible so the user can inspect generation in progress.
13. Keep the private IndexedDB output history with open, delete, and clear-all controls, capped at 20 entries.
14. Keep copy, export, and share actions for generated outputs.
15. Never store raw source notes, raw topics, or raw exam-focus text in local history.
## Validation
Run from this directory:
```sh
npm test
npm run build
npm run check
```
All three commands must pass. The deterministic tests load the production helper script directly from `index.html` and cover protocol parsing, HTML escaping, reference extraction and preservation, formatting, and model allowlist intersection.