← ClaudeAtlas

examforgelisted

Turn a student's own course material into a study guide AND a prediction of what the exam will actually ask, plus an executable manual for how to answer it. Use whenever the user wants to prepare or revise for an exam, midterm, final, practical, OSPE, or quiz from their slides, professor transcripts (English OR Arabic), past exam papers, model answers, lab manuals, or class notes - including "help me study for X", "predict my exam", or uploaded lecture files. Also use when they ask how to solve or answer exam questions: recognizing question types, choosing a method or formula, structuring answers for marks, avoiding the professor's distractors, what to do when stuck, or prioritizing revision with limited hours left. Built for biology-family courses (biotech, biology, biochemistry, bioinformatics, chemistry, microbiology, molecular biology) but adapts to the material given. Do NOT use for writing an exam, cheating during a live exam, or tutoring unrelated to a specific upcoming assessment.
SamDozer/Exam-Forge · ★ 0 · Code & Development · score 70
Install: claude install-skill SamDozer/Exam-Forge
# Exam Study Guide & Predictor Produce two things at once from the student's **own** course material: 1. **A study guide** — organized for fast recall under time pressure. 2. **An exam prediction** — what will most likely be asked, and how, ranked by evidence. The objective is **scoring on this specific exam**, not comprehensive mastery. Prioritize what the *professor* and the *past papers* signal, not what a textbook considers important. ## Core principles - **Source-bound.** Build almost entirely from what the student provides. External knowledge is allowed *only* when a concept is central and repeated across the sources but explained poorly there — and every such addition must be visibly flagged as outside-source (see `references/prediction-and-analysis.md` → "Faithfulness rule"). - **Marks-driven.** Always weight coverage by the mark distribution. A 25-mark topic gets more depth and more predicted questions than a 3-mark one. If you don't know the marks, ask. - **Evidence-weighted prediction.** Predictions are ranked High / Medium / Low confidence, each justified by *which* signals fired. Past exams are the single strongest signal. Full weighting hierarchy in `references/prediction-and-analysis.md`. - **Output in English, but read Arabic.** The professor's transcripts may be Arabic (or code-switched Arabic + English technical terms). Parse them, catch spoken exam cues, and integrate their substance into the English output — attributing them as "professor said (verbal