name: quick_revision
description: Generate ultra-compressed bullet summaries for rapid last-minute revision
Quick Revision Generator
Purpose: Create maximum-density revision notes for last-minute recall.
Format Specification
Structure
## [Topic]
### Core Facts
- Fact 1
- Fact 2
### Key Formulas
- Formula: meaning
### Mnemonics
- [Memory aid]
### Exam Traps
- [Common mistake]
Compression Rules
| Rule |
Example |
| One concept per bullet |
TCP = reliable, connection-oriented |
| Formula + meaning |
O(log n) = halving each step (binary search) |
| Contrast pairs |
Stack = LIFO, Queue = FIFO |
| Acronyms |
ACID = Atomicity, Consistency, Isolation, Durability |
| Pattern recognition |
All ML: data → train → predict → evaluate |
Density Targets
| Topic Size |
Target Lines |
| Single concept |
3-5 bullets |
| Unit/Chapter |
15-25 bullets |
| Entire subject |
50-80 bullets |
Must Include
- Definitions - One-line precise
- Formulas - With variable meanings
- Differences - Between similar concepts
- Examples - One canonical example per concept
- Numbers - Key thresholds, limits, counts
- Traps - Where marks are commonly lost
Must Avoid
- Explanations (use teaching mode if needed)
- Full sentences where phrases work
- Redundant information
- Examples that require context
- Anything that needs re-reading
Output Example
## Quick Revision: Classification Algorithms
### Core
- Decision Tree: splits on best attribute (Gini/Entropy)
- Random Forest: bagging + feature randomness
- KNN: majority vote of k nearest neighbors
- SVM: finds maximum margin hyperplane
- Naive Bayes: assumes feature independence
### Formulas
- Entropy: H = -Σ p log₂(p)
- Gini: 1 - Σ p²
- Accuracy: (TP+TN) / Total
### Traps
- KNN sensitive to k choice and scaling
- Naive Bayes fails with correlated features
- Decision Tree overfits without pruning
By