craft-content-modeling

Craft CMS 5 — Content Modeling

How to structure content in Craft CMS 5. Sections, entry types, fields, Matrix, relations, asset management, and strategic patterns for real projects.

This skill covers content architecture — what goes in the CP, how it's organized, and how templates access it. For extending Craft with PHP (plugins, modules, custom element types), see the craftcms skill.

Companion Skills — Always Load Together

When this skill triggers, also load:

• craft-site — Template architecture, component patterns, routing. Required when content decisions affect how templates render data.
• craft-twig-guidelines — Twig coding standards. Required when writing any Twig examples or template code alongside content modeling.
• ddev — All commands run through DDEV. Required for running project config commands, Craft CLI, and content migrations.

Documentation

• Entries: https://craftcms.com/docs/5.x/reference/element-types/entries.html
• Sections: https://craftcms.com/docs/5.x/reference/element-types/entries.html#sections
• Fields: https://craftcms.com/docs/5.x/system/fields.html
• Field types: https://craftcms.com/docs/5.x/reference/field-types/
• Matrix: https://craftcms.com/docs/5.x/reference/field-types/matrix.html
• Relations: https://craftcms.com/docs/5.x/system/relations.html
• Eager loading: https://craftcms.com/docs/5.x/development/eager-loading.html
• Project config: https://craftcms.com/docs/5.x/system/project-config.html

Use WebFetch on specific doc pages when a reference file doesn't cover enough detail.

The Craft 5 Mental Model

Everything is becoming an entry. Entry types are global (shared across sections and Matrix fields). Fields come from a global pool. This is the "entrification" of Craft — categories, tags, and globals are being unified into entries over a three-version arc:

• Craft 4.4 — entrify CLI commands added to convert categories, tags, and globals to entries
• Craft 5 — Categories, tags, and global sets are deprecated and discouraged for new projects. New category groups, tag groups, and global sets can still be created in the CP (the "New" buttons remain, gated by allowAdminChanges), and existing ones continue to work. A unified "Content" section replaces the fragmented entries view. Custom entry index pages (5.9.0) solve the sidebar organization concern.
• Craft 6 — Categories, tags, and global sets will be removed entirely

For new projects, always use entries: Structure sections for hierarchical taxonomy, Channel sections for flat taxonomy, Singles for site-wide settings. For existing projects, migrate at your own pace using the entrify commands.

Three decisions define your content architecture:

1. Which section type organizes the content (Single, Channel, Structure)
2. Which entry types define its shape (global, reusable across contexts)
3. Which relation strategy connects content together (Entries fields, Matrix, CKEditor nested entries, or a combination)

CMS Editions

Craft CMS has four editions (Solo, Team, Pro, Enterprise) that affect content modeling. The key distinction: if any section needs multiple custom user groups with per-group edit/view restrictions, you need Pro or Enterprise (multiple custom user groups are Pro+ only). Team includes one fixed "Team" group whose permissions are customizable for non-admins, plus a 5-user cap. See references/users-and-permissions.md for the full editions table and permissions architecture.

Choose the edition before modeling — it determines whether you can scope content access by user group, which affects section and field architecture.

Section Type Decision

Section Properties

Beyond the type, sections have settings that matter for content architecture:

• maxAuthors (default 1) — allows multiple authors per entry (new in 5.0.0). Set higher for collaborative content.
• minAuthors (default 1, new in 5.10) — minimum number of authors required to save an entry. Set 1 to enforce that every entry has at least one author. Validated on save; setting minAuthors > maxAuthors is rejected by Craft's section validator. Combine with maxAuthors to define an exact range (min: 1, max: 3 = "1 to 3 authors required").
• enableVersioning (default true) — version history for entries
• defaultPlacement — 'beginning' or 'end' for new entries in structures
• previewTargets — array of {label, urlFormat} objects defining where entries can be previewed. Default: primary entry page. Add custom targets for headless frontends, staging URLs, or PDF previews.

Singles replace globals

Set preloadSingles => true in config/general.php to access singles as global Twig variables by handle — identical to the old globals behavior but with drafts, revisions, live preview, and scheduling.

{# With preloadSingles enabled #}
{{ siteSettings.footerText }}
{{ siteSettings.socialLinks.all() }}

Caveat: Singles always propagate to all sites. This is hard-coded.

Structure queries for navigation

{% set topLevel = craft.entries.section('pages').level(1).all() %}
{% set children = craft.entries.descendantOf(entry).descendantDist(1).all() %}
{% set breadcrumbs = craft.entries.ancestorOf(entry).all() %}
{% set siblings = craft.entries.siblingOf(entry).all() %}

Entry Types in Craft 5

Entry types are defined globally (Settings → Entry Types), then attached to sections and Matrix fields. One entry type can serve multiple contexts.

Key implications:

• Changing an entry type's field layout affects every section and Matrix field using it
• Fields come from the global pool — same field definition reused everywhere
• Per-context name/handle/description overrides available (5.6.0+) — useful when the same entry type serves different purposes in different sections
• The global pool demands careful field naming — use specific handles

Entry Type Visual Identity (Craft 5 new)

Entry types have visual properties that improve the editorial experience:

• icon — custom icon identifier, shown in entry type selectors and Matrix "+" menus
• color — one of 20 options (red, orange, amber, yellow, lime, green, emerald, teal, cyan, sky, blue, indigo, violet, purple, fuchsia, pink, rose, white, gray, black)
• description (5.8.0) — help text explaining what this entry type is for
• group (5.8.0) — collapsible grouping in section/Matrix entry type assignments
• uiLabelFormat (5.9.0) — customize the label shown in element indexes (default '{title}')
• hasTitleField / titleFormat — disable the title field and auto-generate from other fields
• showSlugField (5.0.0) / showStatusField (4.5.0) — hide slug or status from editors
• allowLineBreaksInTitles (5.9.0) — for long-form titles

These settings are configured in Settings → Entry Types and affect all contexts where the entry type is used.

Reserved handles — check every proposed field handle against this list

Craft has 83+ reserved handles across all element types via Field::RESERVED_HANDLES. Validation is case-insensitive. Using any of these as a custom field handle will cause a validation error or silent template collision where the native attribute shadows the custom field.

Before proposing any field handle in a content model, check it against the lists below. When a user asks for a field that matches a native attribute name, always use a domain-specific synonym. For example: user says "I need a title field" — the entry already has a native title, so use heading, headline, or pageTitle depending on context.

Most commonly collided (the ones you'll actually try to use):

Full list of reserved handles shared across ALL element types (from Field::RESERVED_HANDLES):

ancestors, archived, attributes, behaviors, canonical, children, contentTable, dateCreated, dateDeleted, dateLastMerged, dateUpdated, descendants, draftId, enabled, enabledForSite, error, errors, fieldLayoutId, fieldValue, fieldValues, firstSave, icon, id, language, level, lft, link, localized, next, nextSibling, owner, parent, parents, prev, prevSibling, ref, revisionId, rgt, root, searchScore, siblings, site, siteId, siteSettingsId, slug, sortOrder, status, structureId, title, trashed, uid, uri, url, viewMode

Additional per-element-type reserved handles:

Also globally reserved (from HandleValidator::$baseReservedWords, applies to ALL handles): attribute, attributeLabels, attributeNames, attributes, dateCreated, dateUpdated, errors, false, fields, handle, id, n, name, no, rules, this, true, uid, y, yes

Field Instances — Reuse Over Duplication

Fields are defined once globally, then instanced into field layouts. Each instance can override four properties without affecting the global definition:

• Label — different display name per context
• Handle — different template handle per context (5.0.0+)
• Instructions — context-specific help text
• Required — required in one layout, optional in another

This means a single heroImage Assets field can be placed in a Blog Post, a Service Page, and a Project entry type — each with different labels ("Hero Image", "Banner", "Cover Photo") and different required settings. The field definition, type, and settings are shared.

The rule: reuse field definitions via instances. Only create a new field when the type or settings differ (e.g., different allowed volumes, different source restrictions, different character limits). Don't create blogHeroImage, serviceHeroImage, projectHeroImage — create one heroImage and instance it.

But don't reuse just because two fields look alike today. Apply the single-responsibility test (SRP): is there a reasonable scenario where one instance would need different settings — a different character limit, translation method, or allowed volumes? If yes, the similarity is incidental, not essential — make them separate fields. Over-reuse has costs that are hard to reverse later:

• Search can't tell instances apart. The index records keywords against the field, not the instance, so summary::daisy may match a byline instance of the same field (documented). If you rely on distinct field-scoped search, the fields must be separate. (Element-query filtering by the overridden handle — .summary(x).byline(y) — does work; only the search index is instance-blind.)
• A few relational/relatedTo cases are community-reported to need the original handle — verify per case if reused-field results look wrong.

Multi-Instance vs Single-Instance Fields

A field's isMultiInstance() method controls whether it can appear multiple times in the same layout with different handles. This is determined by dbType() — fields that return null are single-instance.

Single-instance caveats: Matrix, Content Block, and Addresses fields CAN be reused across different entry types (placed in multiple field layouts), but they can only appear once per layout and their configuration is fully shared — changing the entry types or settings on a Matrix field affects every entry type using it. Reuse these when contexts genuinely share the same structure. Create separate fields when different contexts need different nested entry types or settings.

Adding Fields to an Entry Type — Reuse-First Workflow

Before proposing any field in a content model change, you MUST:

1. Enumerate the existing field pool. Read config/project/fields/ (or CP Settings → Fields). List every field with its type, handle, and key settings.
2. For each proposed field, classify into one of three categories:
   • Reuse — existing field with matching type and compatible settings. Instance it into the layout; optionally override label, handle, instructions, or required.
   • Reuse with settings review — existing field of the correct type but slightly different settings (e.g., different allowed volumes, different character limit). Propose whether settings can be unified so one definition serves both contexts. Flag this as a decision for the user.
   • Create new — no existing field matches by type or functional purpose. Justify why a new field is needed.
3. For single-instance fields (Matrix, Content Block, Addresses) — if an existing field shares the same nested structure, reuse it. If the contexts need different nested entry types or settings, creating a new field is justified — note the shared-config caveat in your reasoning.
4. Present the plan as a table — one row per proposed field, showing the reuse decision, the existing field being reused (if applicable), and the rationale for any new field. Do not create a new field without this table.

Creating a field that duplicates an existing field's functionality is the most common content modeling mistake — it pollutes the global pool, confuses editors, and makes future refactoring harder.

Entrification — Migrating Legacy Content

CLI Commands

ddev craft entrify/categories <categoryGroupHandle>   # → Structure section
ddev craft entrify/tags <tagGroupHandle>               # → Channel section
ddev craft entrify/global-set <globalSetHandle>        # → Single section

All three accept --section to target an existing section instead of creating a new one. entrify/categories and entrify/tags additionally accept --entry-type and --author; entrify/global-set does not.

As of 5.9.0, these commands are interactive — the handle argument is optional.

What the commands do

• Convert the element type to entries
• Migrate all content and field data
• Convert Categories/Tags fields to Entries fields
• Preserve existing relations
• Assign newly created sections to appropriate entry index pages

Entries field for hierarchical selection

When replacing a Categories field with an Entries field, enable Maintain Hierarchy on the Entries field. This auto-selects ancestors when a nested entry is chosen — replicating the category field behavior.

Tags replacement caveat

For flat taxonomies where editors created terms on-the-fly, the on-the-fly creation UX is not yet available for Entries fields. This is the one area where the legacy Tags field still has a UX advantage. Use a Channel section and accept the two-step workflow (create entry separately, then relate it).

CKEditor vs Matrix vs Content Block

Three tools for structured content within an entry — choose based on editing experience:

• Matrix — reorderable blocks, page builders, data grids, media galleries. Multiple view modes (blocks, cards, cards-grid, index).
• CKEditor with nested entries — rich text with occasional structured content inline (image blocks, code blocks, CTAs). Natural writing flow.
• Content Block (5.8.0) — single non-repeatable nested entry for reusable field groups (SEO metadata, banner config).

For the full decision table, nested entry type patterns, and the CKEditor chunks rendering pattern, read references/content-patterns.md.

After creating sections — tidy the Entries element index

Once Entries sources have been customized (Customize sources), Craft persists the full source list to elementSources.craft\elements\Entry in config/project/project.yaml and appends each newly created section to the bottom with generic columns. That is incomplete shipping of the content model. (If sources were never customized, sections show in default order and there's nothing to tidy — see element-index-sources.md for the precondition.)

Required follow-up (same PR / same change set as the section):

1. Place the source next to role peers under the correct sidebar heading (Indexes, Categories, Includes, etc.).
2. Match tableAttributes and defaultSort to the closest similar section (e.g. FAQ Index to News Index; FAQ Categories to News Categories).
3. Include link only when the section has public URLs; drop empty heading: '' leftovers.

Full checklist, role tables, and column key rules (field: vs fieldInstance:): read references/element-index-sources.md.

Common Pitfalls

• Leaving new section sources at the bottom of the Entries index — after creating sections/singles, tidy elementSources placement and columns to match peers; see element-index-sources.md. An unfinished model is one that ships FAQs without FAQ Index sitting with other Indexes.
• Over-using Matrix — if content needs its own URL, independent querying, or permissions, it should be a separate section with an Entries relation field, not a Matrix block.
• Creating new fields without checking the global pool — before adding any field, run the Reuse-First Workflow above. Enumerate existing fields via config/project/fields/ and default to reusing via instance. The only justification for a new field is a different type or genuinely incompatible settings.
• Over-reusing one field across instances with distinct needs — the opposite mistake. The search index can't distinguish instances of a reused field (summary::daisy may match a byline instance — docs), and a few relatedTo cases need the original handle. Reuse only when the instances are genuinely the same field; if one could reasonably need different settings (single-responsibility), split it. Reuse-first, not reuse-always.
• Vague or reserved field handles — image, text, link are too generic (and link is actually reserved). For every field handle in a content model, follow this check: (1) is the handle in the reserved list? If yes, use a synonym from the table. (2) Is the handle too generic for the global field pool? If yes, add domain context: featuredImage, bodyContent, primaryLink. (3) Don't over-specify — blogFeaturedImage creates a new field when you could instance featuredImage with a label override.
• Not planning multi-site from the start — propagation method, field translation methods, and site settings must be configured before content exists. Changing propagation later resaves all entries.
• Using categories/tags/globals in new projects — they're deprecated and discouraged in Craft 5 (still creatable in the CP) and will be removed in Craft 6. Use entries instead.
• Forgetting preloadSingles — without it, singles aren't available as global variables and you need explicit queries.
• Matrix for everything — 15+ entry types in one Matrix field is a red flag. Deeply nested Matrix hits max_input_vars limits and degrades CP performance.
• Not using .eagerly() — every relational field access inside a loop should use .eagerly() to prevent N+1 queries.
• Editing project config YAML manually — let Craft manage config/project/. Use ddev craft project-config/rebuild to regenerate from DB if needed.
• Using database IDs in URI formats — IDs differ across environments. Use {slug}, {canonicalUid}, or custom fields.
• Not setting allowAdminChanges => false in production — without this, production schema changes won't sync back to dev.
• Using @web in filesystem URLs — @web is auto-detected from the HTTP request and can be spoofed or empty in console/queue contexts. Use environment variables ($ASSETS_URL) for filesystem URLs. @webroot for paths is less risky but env vars are still preferred.
• Ignoring entry type visual identity — editors navigate by icon, color, and description. Investing in these settings makes the CP usable as the content model grows.
• Not planning for CMS edition — if you need per-group content permissions, you need Pro or Enterprise. This affects section and field architecture.

Reference Files

Read the relevant reference file(s) for your task.

Task examples:

• "Add fields to an existing entry type" → Reuse-First Workflow in this SKILL.md
• "Modify an entry type's field layout" → Reuse-First Workflow in this SKILL.md
• "Plan a blog content architecture" → read content-patterns.md
• "Which field type should I use for X?" → read field-types.md
• "Set up relatedTo queries" → read relations-and-eager-loading.md
• "Configure Matrix with nested entries" → read field-types.md (Matrix section)
• "Plan a multi-site content model" → read content-patterns.md + infrastructure.md (propagation)
• "Set up users, groups, and permissions" → read users-and-permissions.md
• "Understand project config workflow" → read infrastructure.md (Project Config Essentials)
• "How does Craft store content internally?" → read infrastructure.md (Storage Model)
• "Set up asset volumes and filesystems" → read infrastructure.md (Assets section)
• "Configure URI format for a structure section" → read object-templates.md (Structure URI Patterns)
• "Set up dynamic asset upload subpath" → read object-templates.md (Asset Subpath Patterns)
• "Asset subpath broken after moving field into Matrix" → read object-templates.md (The Matrix Gotcha)
• "Tidy Entries index after creating FAQ sections" → read element-index-sources.md
• "New section shows at bottom of Entries with wrong columns" → read element-index-sources.md
• "Match FAQ Index / Categories to News Index / Categories sources" → read element-index-sources.md