# Documentation
drft builds a dependency graph from your files and checks it for drift. The
substrate is a **set of independent graphs** — `fs` (a node per file, typed and
hashed), `markdown` (link edges), and `frontmatter` (edges plus metadata). A
composition step merges them by path into one graph, and rules read the composed
graph to emit findings.
## Reference
- [Configuration](config.md) — `drft.toml`: `ignore`, `[graphs.*]`, `[rules.*]`
- [Parsers](parsers/README.md) — how parsers extract links and metadata
- [Rules](rules/README.md) — the drift and structural findings `drft check` emits