Document classifier

Skill cxcscmu/SkillLearnBench/skills/b4-skill-creator-claude-haiku-4-5/organize-messy-files/document-classifier

[COLM'26] SkillLearnBench is the first benchmark for evaluating continual learning methods that automatically generate agent skills.

Install

npx -y skills add cxcscmu/SkillLearnBench --skill document-classifier

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

What its author says it does

Copied from the file, not written here

Classify documents by subject matter. Use this skill when you need to determine which category (LLM, Trapped Ion & Quantum Computing, Black Hole, DNA, Music History) a PDF, DOCX, or PPTX file belongs to based on its content. Analyzes titles, abstracts, keywords, and text content to accurately categorize documents.

SKILL.md

3.2 KB, as published. Nobody here has run it

Document Classifier

Overview

This skill helps classify documents into one of five subject categories by analyzing their content. Each document belongs to exactly one category, so use clear decision criteria to avoid misclassification.

Subject Categories and Keywords

1. LLM (Large Language Models)

Keywords: transformer, BERT, GPT, neural network, language model, NLP, text generation, attention mechanism, embedding, tokenization, fine-tuning, prompt engineering, instruction tuning

2. Trapped Ion and Quantum Computing

Keywords: trapped ion, quantum computing, quantum gate, qubit, quantum algorithm, ion trap, quantum entanglement, quantum error correction, quantum simulation, quantum advantage

3. Black Hole

Keywords: black hole, gravitational singularity, event horizon, Hawking radiation, gravitational collapse, spacetime curvature, general relativity, accretion disk, cosmic string

4. DNA

Keywords: DNA, deoxyribonucleic acid, genetics, genomics, protein, amino acid, mutation, genetic code, sequencing, chromosome, gene expression, CRISPR, molecular biology

5. Music History

Keywords: music, composer, symphony, sonata, classical, baroque, jazz, rhythm, melody, harmony, musical instrument, concert, opera, historical period

Classification Process

Extract content: Read the document title, abstract/summary, and first few paragraphs
Identify keywords: Look for subject-specific terminology from the sections above
Analyze context: Consider the overall theme, methodology, and references
Make determination: Classify into the best-matching category
Handle edge cases: If uncertain between two categories, prioritize based on:
- Primary focus area (main topic vs. secondary mentions)
- Author expertise and publication venue
- Methodological approach (computational vs. theoretical vs. experimental)

File Format Handling

PDF Files

Extract text using PDF parsing (text layer or OCR if needed)
Check metadata (title, author, keywords)
Read abstract and introduction

DOCX Files

Extract all text content
Check document title and any metadata
Read opening paragraphs

PPTX Files

Extract slide titles and content from all slides
Focus on early slides which typically contain overview/subject matter

Decision Boundaries

LLM vs Trapped Ion/QC: LLM focuses on language/text processing; quantum computing on quantum systems Black Hole vs Physics: Black hole is astrophysics-specific; general physics belongs elsewhere DNA vs Biology: DNA is specifically about genetics and molecular biology Music History vs Music Theory: Music history is about historical context and composers; music theory is about structure

Output Format

For each document, provide:

Filename
Determined category (1-5)
Confidence level (high/medium/low)
Key indicators that led to this classification