Install
$ agentstack add skill-refaelach-gmba-class-19-skills-exam-prep ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Exam Prep
The single best signal for what your professor will test, and how, is what they've tested before. Past exams reveal question types, phrasing style, point distribution, time pressure, and the topics that keep coming back. This skill takes one or more past exams from the student, learns the pattern, and produces a fresh mock exam that feels like the real one — with answers withheld until the student actually submits, so it works as practice instead of read-along review.
When to use this skill
- The student has at least one past or sample exam (PDF, image, scan, doc, or pasted text) and wants to practice for an upcoming exam in the same course.
- The student says things like "make me a mock exam from these," "I have last year's final, quiz me," or "what kind of questions should I expect."
- The student has several past exams from the same course and wants to figure out what topics keep repeating (= high-yield) and what the format is likely to be.
- The student wants timed-exam-style practice rather than passive review.
When NOT to use this skill
- The student wants flashcards or a single-topic deep-dive — that's concept review, not exam prep.
- The student wants help solving a specific homework problem — answer it directly.
- The student has zero exam material. Ask for at least one past exam, a problem set, or a syllabus chapter list. Without an artifact, this skill is guessing.
- The student wants the answer key up front. This skill withholds answers until they attempt the exam — that's the point. If they want a worked example, walk them through one without the mock-exam ceremony.
How it works
Step 1: Get the past exam(s)
If the student hasn't already provided exam material, ask for it. Be specific:
> "Drop in any past exams you have — PDF, photos, a Word doc, or even pasted text is fine. One is enough; if you have several from different years, share them all. I'll learn the format from whatever you give me."
If the student has only fragments (e.g., one section, or screenshots of a few questions), say so and proceed — but flag that the mock will be more reliable with a complete exam.
If the student shares an exam that includes the answer key (some profs publish solved past exams), that's fine — use it to learn the expected answer style and depth, but never copy the keyed questions verbatim into the mock and never show the keyed answers before the student submits theirs.
Step 2: Analyze structure and patterns
Read each exam carefully and extract, for the student to see, a brief profile:
- Question types present: multiple choice, true/false, short answer, definitions, fill-in-the-blank, calculation/numerical, case scenario, essay/long-form, compare-and-contrast, etc. Note the rough mix (e.g., "60% MCQ, 25% short answer, 15% case essay").
- Structure: number of sections, how each section is labeled, point distribution across sections, total points, time limit if shown.
- Phrasing style: formal vs. casual, command verbs the prof favors ("Discuss," "Compute," "Critique," "Compare with reference to…"), whether questions assume specific frameworks or terminology.
- Difficulty mix: how much is recall, how much is application, how much is judgment / open-ended analysis.
- Topic coverage: list the topics each question hits. If multiple exams are provided, highlight what repeats — topics that show up in two or more exams are high-yield and should be over-represented in the mock.
- Format constraints if mentioned: open book, formula sheet allowed, calculator policy, language requirement.
Give the student a quick 5–8 line summary of what you found before generating the mock. This both confirms the pattern was read correctly and helps them notice things ("oh, every year there's a forecasting question — I should focus on that").
Step 3: Generate the mock exam
Produce a complete mock exam that mirrors the structure of the originals. Specifically:
- Same number of sections, same section labels, same point distribution.
- Same question type mix and roughly the same number of each type.
- Same phrasing style and command verbs as the originals — match the formality and the way the prof tends to ask things.
- Same difficulty mix.
- New questions, not paraphrases of the originals. The questions should cover the same kinds of topics the originals covered, with extra weight on topics that repeated across multiple past exams.
- Show the time limit at the top if the originals had one. Show point values per question if the originals did.
- If the originals declared format constraints (open book, formula sheet, calculator), restate them at the top so the student knows the conditions to simulate.
End the mock with clear instructions for the student:
> How to take this: Try to do this under realistic conditions — same time limit, same allowed materials. Write or type your answers. When you're done (or you've gotten as far as you can), tell me you're ready and I'll go through every question with you: correct answer, how to get there, and what the question was testing. I won't show any answers until you submit. If you want to skip a question, say "skip Q3" and I'll come back to it. If you want a hint without the full answer, say "hint on Q3" and I'll nudge without spoiling it.
Step 4: Wait for the student to take the exam
This is the part that makes this skill actually useful, so honor it:
- Do not reveal answers, hints, or solution methods before the student attempts. If they ask "what's the answer to Q4," respond with "take a swing at it first — I'll grade as soon as you submit. Want a hint without the full answer?"
- A "hint" should nudge toward the right framework or formula without doing the work. E.g., "Q4 is testing your understanding of NPV under uncertainty — what does that imply about the discount rate?"
- If the student says "skip Q3," mark it skipped and continue waiting on the rest. Don't grade Q3 until the rest are submitted, in case they come back to it.
- The student signals they're done by saying something like "done," "I'm ready," "submit," "show me the answers," "grade it." Treat any of those as the trigger to move on.
Step 5: Reveal answers and explanations
For each question, in order, give the student:
- The correct answer (or model answer for open-ended questions).
- How to get there — the reasoning, formula, framework, or solution path. Not just the answer; the method.
- What the question was testing — the underlying concept, framework, or skill. This is what helps the student generalize.
- (For MCQ/T-F) a brief note on why the wrong options are wrong. Distractors are pedagogical — the student should know why each one looked tempting.
- (For essays / long-form) a rubric or checklist of the points a strong answer should hit, plus a brief assessment of what they wrote against that rubric — what they got, what they missed. Be honest. False praise wastes their study time.
After grading, give a short overall debrief:
- Which question types they handled well and which they struggled on.
- Which topics they should drill more (with specifics, not "review more").
- One or two highest-leverage things to fix before the real exam.
If the student wants more practice on a weak area, offer to generate a focused drill on that question type or topic next.
Inputs the skill expects
- At least one past or sample exam — PDF, image, Word doc, or pasted text. More is better; repetition across years signals high-yield topics.
- Optional: course name, exam date, topics to emphasize.
- Optional: format constraints (time limit, open-book, calculator policy). If silent, mirror the originals.
Outputs the skill produces
- A short structural profile of the past exam(s) — the patterns the skill detected.
- A complete mock exam, formatted to mirror the originals (sections, points, time, instructions, question style).
- After the student submits: per-question grading with correct answer, solution method, what each question tested, and (for essays) a rubric-based assessment of what they wrote.
- A short debrief at the end with concrete weak-area drilling suggestions.
Example
Student: "Corporate Finance final next Tuesday. Here are the last three years' finals — can you make me a mock?"
The skill:
- Reads all three. Detects the pattern: 2 hours, 90 points, three sections (10 MCQs × 2 pts, 3 numerical × 10 pts, 1 case essay × 40 pts). Formula sheet + financial calculator allowed. WACC appears in all three years, NPV/IRR in two, capital structure in one.
- Reports the pattern back to the student so they can confirm, and notes it'll over-weight WACC and NPV since they keep recurring.
- Generates a fresh mock in the same shape — new numbers, new scenarios, same structure. The case essay asks them to evaluate a financing recommendation for a fictional firm.
- Sets exam conditions and waits — no answers shown until submission.
- After "done," grades each item: MCQs with distractor analysis, numericals with full solution paths, case essay against a rubric (correct WACC? right tradeoffs? defensible recommendation?).
- Closes with a debrief: "MCQs strong (9/10). You lost ground on the WACC numerical — used book weights instead of market. Drill three more before Tuesday. Case essay was solid but light on quantitative support — cite specific numbers next time."
Notes for contributors
- The "withhold answers until submission" rule is the load-bearing part of this skill. Without it, the skill collapses into a study guide that the student reads instead of practicing on. Keep that rule even if it feels like friction.
- If the student is short on time and asks to skip the practice phase ("just show me what kinds of questions to expect"), it's OK to bail out of the full mock-exam flow and instead produce an annotated outline: question types, topics, point distribution, sample stems. Don't force the full exam if it's not what they need right now.
- For multi-exam input, repetition across years is the strongest signal of what matters. Surface it explicitly to the student — they may not have noticed.
- Imagery / scanned exams sometimes parse imperfectly. If a question is partly unreadable, say so and ask the student to type that question rather than guessing.
- Idea for v2: a "diagnose first" mode where the skill picks five short representative questions from the past exams as a 15-minute diagnostic, identifies weak spots from how the student does, and then generates a mock weighted toward those weak spots.
Author: Refael (Rafa) Lachmish (@refaelach) Created: 2026-04-25 Tested on: Claude (Opus 4.7) Real tasks tested: Uploaded a past Statistics exam and asked the skill to generate practice questions. The skill detected the question types in the original exam and produced custom practice questions that matched those types.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: refaelach
- Source: refaelach/gmba-class-19-skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.