vm-document-section-classifier

# VM Document Section Classifier

## Overview

Given a document converted to markdown, this skill classifies each section by content type and determines its suitability as an evidence source. Classification drives downstream processing — agents use the classification to filter which sections to search for requirements or capability evidence, and which to skip (boilerplate, marketing, preamble).

This skill is generative: it interprets section content semantically to assign classifications. Ambiguous sections are flagged for human review.

## Brownfield Context

In brownfield scenarios, documents being analyzed often contain mixed content — a system appreciation document may include executive summaries, implementation notes, policy statements, and marketing claims alongside actual functional specifications. This skill separates the signal from the noise before requirement extraction or gap analysis begins.

## When to Use

- Before extracting requirements from a system appreciation document (filter preamble from specification content)
- Before mapping stated capabilities to a codebase (identify functional vs marketing sections)
- Before cross-referencing a change request against RE Agent output (focus on scope/impact sections)
- Any workflow where a large document needs content-type triage before processing

## When NOT to Use

- Document is already a structured SRS with clearly separated requirement IDs
- Single-section documents with no meaningful subdivision
- Non-text content (use vm-document-markdown-converter first)

## Inputs

```json
{
  "document": {
    "path": "path/to/document.md",
    "title": "Document title",
    "source_format": "pdf|docx|html|markdown"
  },
  "taxonomy": "requirements|system_appreciation|standards|specification|custom",
  "custom_taxonomy": {
    "categories": {},
    "evidence_rules": {}
  }
}
```

## Built-in Taxonomies

### Requirements Document Taxonomy (`taxonomy: "requirements"`)

| Classification | Description | Evidence Role |
|---|---|---|
| `functional_requirement` | Feature/behavior specifications with "shall/must/will/should" language | PRIMARY — extract as requirements |
| `non_functional_requirement` | Performance, security, scalability, reliability constraints | PRIMARY — extract as NFRs |
| `scope_boundary` | What is in/out of scope for the change or system | PRIMARY — extract as constraints |
| `assumption` | Stated assumptions about the environment or users | SECONDARY — flag for validation |
| `preamble` | Introduction, document purpose, revision history | EXCLUDED |
| `procedural` | Process instructions for document use | EXCLUDED |

### System Appreciation Document Taxonomy (`taxonomy: "system_appreciation"`)

| Classification | Description | Evidence Role |
|---|---|---|
| `capability_description` | What the system currently does — behavior, features, flows | PRIMARY |
| `integration_description` | External systems, APIs, data feeds connected to this system | PRIMARY |
| `data_description`