---
title: "Phase 3: Decorating the Data with Subagents"
description: Analyze converted content to build a taxonomy schema, create a classification agent, and run parallel classification at scale.
---

# Phase 3: Decorating the Data with Subagents

Now that you have 209 clean markdown files, it's time to add meaning. This phase covers the most sophisticated patterns: analyzing content to design a taxonomy schema, building a specialized classification agent, and running parallel classification at scale.

The key patterns: **Semantic Enrichment** (LLMs extracting meaning from content), **Agents Writing Agents** (describing the classifier in plain language), and **Parallel Execution** (processing 50 files simultaneously).

## Lessons

1. **[Taxonomy Design](/content-repurpose/03-classification/06-taxonomy-design/)** — Analyze all content files to identify patterns, series, and topics. Design a comprehensive frontmatter schema.

2. **[Parallel Classification](/content-repurpose/03-classification/07-parallel-agents/)** — Create a classification agent and run it against batches of files with concurrent subagents.

3. **[Semantic Enrichment](/content-repurpose/03-classification/08-semantic-enrichment/)** — LLMs extract people, companies, models, audience level, and content depth — metadata that would be tedious to create manually.