> ## Documentation Index
> Fetch the complete documentation index at: https://mehen.ophi.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Section tree

> Heading-derived section structure with quality flags (skips, chunking smell, fragmentation smell).

Headings produce a derived **section tree** that downstream metrics ([MCC](/metrics/markdown/mcc),
[Filler / Lazy Risk](/metrics/markdown/filler-lazy-risk),
[Section Balance](/metrics/markdown/section-balance)) read from. Every section has a heading level,
byte/line range, parent/child IDs, word count, and per-section artifact and link counts.

## Structure

Each section node carries:

```text theme={null}
Section
├── heading_level   : 1..6
├── byte_range
├── line_range
├── parent_id, children_ids
├── word_count
├── artifact_counts : { code, table, image, diagram, math }
└── link_counts     : { internal, relative, external, broken }
```

## Quality flags

mehen marks three structural smells on the section tree:

| Flag                    | Trigger                                                                    |
| ----------------------- | -------------------------------------------------------------------------- |
| **Heading skip**        | Direct nesting jump (e.g., `H1 → H4` without an intervening `H2`/`H3`).    |
| **Chunking smell**      | An `H2` section that is oversized relative to the document's `p95` length. |
| **Fragmentation smell** | A document flooded with very short `H5`/`H6` nodes — over-decomposition.   |

These flags surface on the [PR comment](/guides/pr-comment-design) as advisory callouts.

## How downstream metrics use the tree

* [**MCC**](/metrics/markdown/mcc) reads heading-level changes to charge nesting penalties.
* [**Filler / Lazy Risk**](/metrics/markdown/filler-lazy-risk) reads section sizes to detect "shallow big
  doc".
* [**Section Balance**](/metrics/markdown/section-balance) penalizes oversized sections at the 95th
  percentile.
* [**Evidence Coverage**](/metrics/markdown/evidence-coverage) computes per-section evidence anchor
  density.

## References

* W3C: [Web Content Accessibility Guidelines 2.1 — Heading levels](https://www.w3.org/WAI/WCAG21/Understanding/headings-and-labels.html)
  — the source of the heading-skip quality flag (assistive-technology users rely on heading-level
  ordering).
* CommonMark Spec: [ATX headings](https://spec.commonmark.org/0.30/#atx-headings) and
  [setext headings](https://spec.commonmark.org/0.30/#setext-headings) — the parser-level definition
  the section tree builds on.

## See also

* [LOC family](/metrics/markdown/loc-family) — feeds per-section line counts.
* [Section Balance](/metrics/markdown/section-balance) — balance scoring on the tree.