Imported from pinakdhabu/Exam-prompt (skills/qp-fetcher/SKILL.md). Install upstream with npx skills add pinakdhabu/Exam-prompt --skill qp-fetcher. Copyright stays with the author.

Universal QP Fetcher & Generator

Overview

This skill provides two capabilities:

1. Fetch: Discovers and structures real question papers via LLM web search (primary) or automated download scripts (fallback for SPPU)
2. Generate: Creates university-style question papers with proper formatting, instructions, OR structure, CO mapping, Bloom's levels

Works for ANY university worldwide, ANY branch, ANY semester.

How It Works

Option A (PRIMARY): LLM Web Search → Structured Markdown

Use the LLM's web search capability to discover question papers for any subject/university. This is the recommended approach — no single-site dependency, works for any university globally.

Process:

1. LLM searches the web for question papers matching: "{subject}" "{university}" question paper {year}
2. Also searches university portals, educational sites, and PYQ repositories
3. Extracts actual question text from search results (no PDF downloads needed)
4. Writes structured markdown to the project's PYQ index

Output format (pyq-index/{university-slug}/{subject-slug}.md):

---
subject: Database Management Systems
university: Savitribai Phule Pune University
pattern: 2019
collected: 2026-06-22
sources:
  - type: web
    url: https://example.com/pyq/dbms-2024
  - type: web
    url: https://example.com/pyq/dbms-2023
---

# DBMS — Previous Year Question Papers (SPPU 2019 Pattern)

## End Semester Examination

### April 2024

**Total Marks: 70 | Time: 2.5 hrs**

**Q.1** Explain DBMS architecture with neat diagram. (10M) **Q.2** a) What is normalization? Explain
3NF and BCNF with examples. (8M) b) Write SQL queries for given relational schema. (7M) ...

### November 2023

...

## In Semester Examination

### October 2023

**Total Marks: 30 | Time: 1 hr**

**Q.1** ...

Search strategy per university type:

If LLM Web Search Fails (any non-SPPU university)

If the LLM cannot find enough PYQs on the web after reasonable searching, ask the user for:

Then retry the original task with the user-provided material.

SPPU Internet Sources

These are the known working sources for SPPU question papers. LLM should search these in order until enough papers are found.

⚠️ Links change every exam season. SPPU creates new folders per exam period on all sites. If a URL 404s, try the next source or search for the current exam period's folder. Always update these instructions when you discover new URLs or find broken ones.

Official SPPU Portal — Dynamic URL Discovery

Don't hardcode folder names. They change every exam season. Instead, discover dynamically:

http://collegecirculars.unipune.ac.in/sites/examdocs/

Step 1 — Browse the site structure: Fetch the base URL above. The SharePoint sidebar lists all available libraries. Look for:

• Archive Question Papers section — these are the QP folders
• Time Tables {SEASON} {YEAR} — timetable folders (each exam season gets a new one)
• Each archive folder is named after its exam period (e.g., APRIL - 2025, November - 2024)

Step 2 — Derive the folder URL pattern:

http://collegecirculars.unipune.ac.in/sites/examdocs/{FOLDER_NAME}/Forms/AllItems.aspx

URL-encode spaces in folder names. Common encoding patterns seen:

• APRIL - 2025 → APRIL%20%202025 (double space after month)
• November - 2024 → November%20-%202024
• April-2024 → April2024 (no encoding needed)

Try multiple encodings if the first fails. The site is inconsistent year-to-year.

Step 3 — Download PDFs using the known pattern: Each period folder contains merged all-branch PDFs named by year level:

{PERIOD_URL}/{YEAR_LEVEL} ({PATTERN} PATTERN).pdf

• YEAR_LEVEL = F.E, S.E, T.E, or B.E
• PATTERN = 2019 PATTERN or 2024 PATTERN

Try: S.E%20(2019%20PATTERN).pdf, T.E%20(2019%20PATTERN).pdf, B.E%20(2019%20PATTERN).pdf, F.E%20(2019%20PATTERN).pdf, F.E%20(2024%20PATTERN).pdf, etc.

Step 4 — Timetables (same dynamic approach):

http://collegecirculars.unipune.ac.in/sites/examdocs/Time%20Tables%20{SEASON}%20{YEAR}/Forms/AllItems.aspx

Current/known patterns: APRMAY, OCT/NOV, MAR/APR + year. Browse sidebar to find latest.

Limitations of official portal:

• Only End Semester papers (no In Sem)
• PDFs are large merged files (all branches in one PDF — use Ctrl+F to find your subject)
• Requires navigating a SharePoint UI (not always easy to scrape)
• Folder naming is inconsistent (spaces, hyphens, URL encoding vary per year)

LLM process for each source:

1. Search the URL with the subject name
2. If the page lists individual paper links, open each link and read the PDF directly
3. Extract question text, marks, exam type from the PDF
4. If direct PDF reading fails, search for the same paper on another source
5. Write collected questions as structured markdown

Branch slug mapping for sppuquestionpapers.com:

computer-engineering, entc, mechanical-engineering, civil-engineering,
electrical-engineering, information-technology

Semester URL pattern:

https://sppuquestionpapers.com/be/{branch}/semester-{N} (N = 1-8)

Option B (FALLBACK): Automated Scripts (SPPU only)

If web search fails or SPPU-specific batch downloads are needed, use the automated fetcher:

# Fetch all available DBMS question papers
node scripts/fetch-qp.js dbms

# Fetch DBMS papers from 2024 only
node scripts/fetch-qp.js dbms --year 2024

# Fetch all Semester 5 Computer Engineering papers
node scripts/fetch-qp.js --semester 5

SPPU Hyperawareness — All Official Portals

For SPPU, the LLM must be aware of all these portals and use them dynamically. Links change every exam/syllabus season — always discover dynamically by browsing the base site.

Base domains:

• http://collegecirculars.unipune.ac.in/ — Timetables, exam docs, syllabi, academic calendar
• http://onlineresults.unipune.ac.in/ — Result checking portals
• https://unipune.ac.in/ — Main university website
• https://exam.unipune.ac.in/Pages/PreviousQuestionPapers.html — Alternative QP official source
• https://sppuresult.vercel.app/ — Community result viewer (fetches live from SPPU)

Syllabus archives (dynamic discovery): Browse http://collegecirculars.unipune.ac.in/sites/documents/ sidebar → Syllabus {YEAR} folders. Pattern: http://collegecirculars.unipune.ac.in/sites/documents/Syllabus%20{YEAR}/Forms/AllItems.aspx or .../Syllabus{YEAR}/Forms/AllItems.aspx (encoding varies). Available years: 2016–2026.

What to do when user asks about SPPU:

1. Check pyq-index/ for existing markdown first
2. If QPs needed → use dynamic discovery on official portal (collegecirculars.unipune.ac.in)
3. If results needed → use sppuresult.vercel.app or onlineresults.unipune.ac.in
4. If syllabus needed → browse collegecirculars syllabi folders
5. If timetables needed → browse Time%20Tables%20{SEASON}%20{YEAR}/ on examdocs
6. If revaluation needed → unipune.ac.in/university_files/Reval_Online_Results_online.htm
7. If academic calendar needed → collegecirculars.unipune.ac.in/sites/documents/Academic%20Calender/
8. If circulars needed → collegecirculars.unipune.ac.in/SitePages/Home.aspx
9. If QP alternative needed → exam.unipune.ac.in/Pages/PreviousQuestionPapers.html
10. Always write back to pyq-index/ for future offline use

⚠️ All URLs above can change. If a link fails, browse the parent site to find the new location.

For other universities:

Same approach — LLM dynamically discovers portals by searching: "{university}" syllabus, "{university}" exam timetable, "{university}" result portal, "{university}" question papers, "{university}" academic calendar If not found → ask user for links/files.

Option B: Generate University-Style Question Paper

When the user says "Generate a question paper for [subject]", the AI generates a question paper matching the Real QP Style Reference section below (PDF analysis of 200+ actual SPPU QPs):

1. Page Layout: A4, margins ~70pt left/right, ~56pt top, ~60pt bottom (matching real QPs)
2. Header: Paper code + SEAT No. + pages on first line, centered paper number/program/subject
3. Exam Details: Time on left, Max Marks on right, proper instruction set as seen in real QPs
4. Question Structure: OR pairs with (a)+(b)+(c) format, marks in brackets
5. Mark Distribution: Unit-wise as per SPPU pattern
6. CO/Bloom Mapping: Each question tagged with CO and Bloom's level
7. Typography: ALL CAPS for subject, brackets for marks, bold for headings

Supported Patterns

Website Structure

The QP fetcher scrapes question papers from sppuquestionpapers.com:

https://sppuquestionpapers.com/
├── /be/                         # Bachelor of Engineering
│   ├── /computer-engineering/
│   │   ├── /semester-3          # SE Sem III papers
│   │   ├── /semester-4          # SE Sem IV papers
│   │   ├── /semester-5          # TE Sem V papers
│   │   ├── /semester-6          # TE Sem VI papers
│   │   ├── /semester-7          # BE Sem VII papers
│   │   └── /semester-8          # BE Sem VIII papers
│   ├── /entc/
│   ├── /mechanical-engineering/
│   ├── /civil-engineering/
│   ├── /electrical-engineering/
│   └── /information-technology/
├── /me/                         # Master of Engineering
├── /mca/                        # MCA
├── /bsc/                        # B.Sc.
└── /bcom/                       # B.Com.

Each page lists subjects with Year, Month, and Download buttons.

Subject Aliases

The fetcher supports fuzzy subject matching:

Question Paper Generation Format

When generating a question paper (not fetching), replicate the exact visual style of real SPPU question papers. The generated markdown should be structured so that when rendered to PDF, it matches the official layout properties documented below.

Real QP Style Reference (from 200+ VL + FE Computer Engineering PDFs)

These properties were extracted via pdfminer/pdf2txt.py + pdftotext analysis of representative samples across FE/SE/TE/BE and 2012/2015/2019/2024 patterns. Use this as the authoritative style guide.

Page Properties:

Note on FE 2024 Pattern: Normalized to A4 (same layout as 2019 SE/TE/BE pattern) for generation. Original source PDFs used Letter (PageMaker) or non-standard 560×792 pt (cairo), but A4 production is the target. Paper code format varies: PC-#### / [6351]-XXX (PageMaker variant, Nov 2024) or PD#### / [6401]-XXXX (cairo variant, April 2025).

Common Layout Elements (all patterns):

[Left-aligned header]                           [Right-aligned header]
Total No. of Questions : N]                     SEAT No. : [number]
[Paper Code]                                    [Total No. of Pages : N]

           [Paper Number] e.g., [6352]-35
           [Program Name] e.g., S.E. (Computer Engineering)
           [SUBJECT NAME IN ALL CAPS]
           (Pattern) (Semester) (Subject Code)

Time : [X] Hours]                               [Max. Marks : XX]

Instructions to the candidates:
  1) [Instruction 1]
  2) [Instruction 2]
  ...

[Question body with marks in brackets]

Typographic Rules:

• Paper number like [6352]-35 — bold, centered, larger than body
• Subject name — ALL CAPS, bold
• Pattern/semester info — in parentheses, regular weight
• Time and marks — bold, same line, left/right justified
• Instructions — numbered with ), indent 3 spaces
• Question numbers — bold like **Q1)** or **Q.1**
• Sub-question labels — lowercase a) b) c), bold, indent
• Marks — in [N] brackets after question text, right-aligned or inline
• OR separator — **OR** centered on its own line

Mark Structure (SPPU 2019 Pattern):

• End Sem: 70 marks, 2.5 hrs, 8 questions (4 OR pairs), single compulsory at Q.1
• In Sem: 30 marks, 1 hr, 3 questions (all compulsory)
• Each question: 2–3 sub-questions with marks [2]–[8] each

Question body format:

**Q1)** State the following with justification.                      [6]
    a) [statement]                                                   [2]
    b) [statement]                                                   [2]
    c) [statement]                                                   [2]

**OR**

**Q2)** a) [question text]                                            [6]
        b) [question text]                                            [6]
        c) [question text]                                            [6]

Generation Templates

Use the following markdown templates when generating question papers. The template aligns with the official PDF layout above.

Outer box header (for visual rendering or markdown-to-PDF conversion):

╔══════════════════════════════════════════════════════════════╗ ║ SAVITRIBAI PHULE PUNE UNIVERSITY
║ ║ ║ ║ [Subject Code]: [Subject Name] ║ ║ [Program] ([Pattern]) ([Semester]) ║ ║ ║ ║ Time: [X]
Hours Total Marks: [XX] ║ ╚══════════════════════════════════════════════════════════════╝

Inner header (for inclusion in a larger document):

Total No. of Questions : [N] SEAT No. :

         **[Paper Number]**
         **[Program Name]**
         **[SUBJECT NAME]**
         ([Pattern]) ([Semester]) ([Subject Code])

Time : [X] Hours] [Max. Marks : XX]

**Instructions to the candidates:**

1. [instruction]
2. [instruction]

Question body (8 questions, 4 OR pairs — 2019 pattern):

**Q1)** [Question text] [M] a) [sub-question] [m] b) [sub-question] [m] c) [sub-question] [m]

**OR**

**Q2)** a) [sub-question] [m] b) [sub-question] [m] c) [sub-question] [m]

**Q3)** ... (same pattern)

...through Q8.

Question body (10 questions, 5 OR pairs — 2024 pattern): Same as above but 5 OR pairs (Q1/Q2, Q3/Q4, Q5/Q6, Q7/Q8, Q9/Q10).

Question body (10 questions, 5 OR pairs — 2015 pattern): 10 questions, 5 OR pairs, 3 hrs duration. Same format.

Question body (8 questions, 4 OR pairs — 2012 pattern): 8 questions, 4 OR pairs, 2.5 hrs. Note: SE 2012 may have different formatting; verify against actual PDFs if available.

Session Config

This skill integrates with the session config system (deps/session-profile.json). Before executing, check for an existing session profile:

• If deps/session-profile.json exists, read university, subject, pattern, and exam_type fields to auto-configure the skill.
• If the file does not exist, fall back to user-provided context or prompt the user to run setup-exam-prompt (or npm run init) first.
• Session config eliminates redundant context detection — detection happens once and is reused across all skill calls.

Integration with Other Skills

Fair Use & Copyright Notice

Question papers are educational material. When fetching from public university portals:

1. Use fetched papers only for personal/educational study purposes.
2. Do not republish fetched papers as your own content.
3. Respect the university's terms of service for their question paper portal.
4. If the paper has a visible copyright notice, honour it.
5. When generating synthetic question papers (not fetched), they are original works — no copyright concerns.

Rate Limiting & Polite Scraping Rules

• Add a delay of 1–3 seconds between successive requests to the same university portal.
• Do not make more than 10 requests per minute to any single domain.
• Check robots.txt if fetching from a root domain (not university-specific portals).
• If a request returns 429 (Too Many Requests) or 503 (Service Unavailable): back off for 30 seconds, retry once. If it fails again, use generation fallback.
• For SPPU portal: limit concurrent downloads to 3. Portal latency ~800ms typical.

Non-SPPU Source Templates

For universities without a dedicated script, use this general approach:

1. Search web: "[University Name] [Subject] [Year] question paper"
2. Search repository: "site:[university-domain].ac.in question paper [subject]"
3. Search PDF aggregators: "site:pdfslide.net [subject] question paper [university]"
4. If no results: fall back to generation rather than fetching nothing.

Fallback: User Upload

When all fetch methods fail for a university:

1. Politely ask: "I couldn't find question papers for [university] automatically. Could you upload 2–3 PDFs of past question papers for [subject]? I can continue with analysis once you share them."
2. If the user uploads a PDF, route to universal-document-reader for text extraction, then proceed with analysis.
3. If the user cannot provide any PDF, offer to generate a realistic question paper in the university's estimated pattern based on the available context.

Prerequisites

• Node.js 18+ (primary runtime)
• Playwright + Chromium (npx playwright install chromium)
• Internet connection (for fetching)
• Run bash scripts/setup.sh to install all dependencies automatically