document-classification
Automates the classification of financial documents like insurance and pension statements.
Install
mkdir -p .claude/skills/document-classification && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/14494" && unzip -o skill.zip -d .claude/skills/document-classification && rm skill.zipInstalls to .claude/skills/document-classification
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Use when working on document classification — `classify_document` tool, OCR pipeline, document type taxonomy for insurance/pension documents.Key capabilities
- →Classify documents using an LLM
- →Define document type taxonomy
- →Design OCR to classification workflows
- →Write tests for document classifiers
- →Log classification decisions
- →Flag low-confidence classifications for review
How it works
The skill classifies documents using the `classify_document` tool, which employs a Claude Haiku model. It assigns a document type, confidence score, and flags items for human review if confidence is below 0.75.
Inputs & outputs
When to use document-classification
- →Building a document classification pipeline
- →Training/testing document classifiers
- →Defining document taxonomy
About this skill
Document Classification skill
When to invoke
- Touching
src/tools/classify-document.ts - Defining or extending document type taxonomy
- Designing OCR → classification flow
- Writing tests for the classifier
Document taxonomy (initial)
| Type | Hebrew | Triggers |
|---|---|---|
id_card | תעודת זהות | "תעודת זהות", ID number patterns |
pension_statement | דוח פנסיה | קרן פנסיה, יתרה צבורה |
gemel_statement | דוח גמל | גמל, השתלמות |
policy_document | פוליסה | פוליסה, ביטוח מנהלים |
loan_document | מסמך הלוואה | הלוואה |
severance_form | טופס פיצויים | פיצויים, סיום עבודה |
salary_slip | תלוש שכר | תלוש, שכר |
medical_form | טופס רפואי | הצהרת בריאות |
power_of_attorney | ייפוי כוח | ייפוי כוח |
unknown | לא ידוע | (fallback) |
Classification policy
- Use Claude Haiku (cheap, fast) — model
claude-haiku-4-5-20251001 - Confidence threshold: if classifier confidence < 0.75 → return
unknown+ flag for human - Always preserve original file path
- Log every classification decision (input snippet, output type, confidence, model)
- Never throw away the original — store classification metadata alongside
Output schema
{
document_type: DocumentType;
confidence: number; // 0-1
reasoning: string; // why this type
hebrew_label: string; // for UI/CRM
requires_review: boolean; // true if confidence < threshold
}
How it compares
This skill automates document classification with an LLM and predefined taxonomy, unlike manual sorting or rule-based systems.
Compared to similar skills
document-classification side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| document-classification (this skill) | 0 | 2mo | No flags | Intermediate |
| docetl | 2 | 2mo | Review | Intermediate |
| slm-lab-benchmark | 1 | 5mo | Review | Advanced |
| reasoning-judge | 0 | 3mo | No flags | Advanced |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
docetl
ucbepic
Build and run LLM-powered data processing pipelines with DocETL. Use when users say "docetl", want to analyze unstructured data, process documents, extract information, or run ETL tasks on text. Helps with data collection, pipeline creation, execution, and optimization.
slm-lab-benchmark
kengz
Run SLM-Lab deep RL benchmarks, monitor dstack jobs, extract results, and update BENCHMARKS.md. Use when asked to run benchmarks, check run status, extract scores, update benchmark tables, or generate plots.
reasoning-judge
kamaz66137-byte
推理判断技能：二分类判断、多分类判断、阈值判定、置信度评估、异常检测、排序判断。触发场景：当用户提到"判断"、"分类"、"是否"、"评估"、"异常"、"judge"、"classify"、"threshold"、"confidence"、"anomaly"时加载。
image-analysis
ComeOnOliver
图片分析与识别，可分析本地图片、网络图片、视频、文件。适用于 OCR、物体识别、场景理解等。当用户发送图片或要求分析图片时必须使用此技能。
adding-a-model
iopsystems
Use when adding a new model (or model family) to the llm-calc database (src/data/models.ts). Walks through sourcing fields from HuggingFace config.json, reading papers for novel attention variants, writing tests first, and registering the entry. Invoke this whenever a model is being added or updated
tss-pipeline
LocNguyen-247
Use when implementing or debugging the TSS remote-sensing workflow in this workspace: Landsat/Sentinel preprocessing, ACOLITE atmospheric correction, cloud/water masking, adjacency correction, station matchup, and model training.