create-classifier
Train task-specific classifiers for the extractor pipeline. Supports vision, text, and hybrid classifiers with GRPO training and execution feedback. Includes data collection, confidence-based routing, and shadow deployment.
npx skills add majiayu000/claude-skill-registry --skill create-classifier --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
# Create Classifier Skill > **Purpose**: Training infrastructure for task-specific classifiers that improve extractor pipeline accuracy through ML-based detection instead of regex/heuristics. ## Overview This skill provides end-to-end infrastructure for creating, training, and deploying classifiers for extraction tasks: - **Data Collection**: Mine labeled data from successful pipeline runs - **Training Templates**: Vision, text, and hybrid classifier architectures - **Execution Feedback**: GRPO-style training with pipeline success as reward - **Confidence Routing**: Automatic fallback to heuristics when confidence is low - **Shadow Deployment**: Compare classifier vs heuristics before full rollout ## Success Story: Table Strategy Classifier The table extraction classifier (S05) achieved: - **95.07% accuracy** (vs ~75% heuristic baseline) - **Reduced fallback rate** from ~25% to <10% - **Faster inference** than multi-strategy attempts This skill generalizes that success pattern for other extraction tasks. --- ## Supported Classifier Types ### 1. Vision Classifiers **Use case**: Document-level classification from images - **Example**: S00 document type detection (arxiv, requirements_
- Overview
- Success Story: Table Strategy Classifier
- Supported Classifier Types
- 1. Vision Classifiers
- 2. Text Classifiers
- 3. Hybrid Classifiers
- Usage
- Step 0: Preflight Assess (Recommended)
- Step 0b: Benchmark-First Model Selection (Recommended)
- Step 1: Data Collection
- Step 2: Train Classifier
- Step 3: Evaluate
- Step 4: Shadow Deploy
- Step 5: Deploy
Supervised fine-tuning (SFT) GRPO with execution feedback Iterative train with preflight assess, optional HF augmentation, and holdout gate Strict quality gate profile (recommended default) Run training in a container (recommended for reproducibility):
What does the create-classifier skill do?
Train task-specific classifiers for the extractor pipeline. Supports vision, text, and hybrid classifiers with GRPO training and execution feedback. Includes data collection, confidence-based routing, and shadow deployment.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill create-classifier --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.