CSS Counters
CSS counters are document-scoped variables maintained by the rendering engine for automatic numbering — headings, figures, steps, and nested outlines — without hard-coding numbers in HTML.
Use counter-reset, counter-increment, counter()/counters() in content, and @counter-style for custom systems.
info
| 1 | article { counter-reset: section figure; } |
| 2 | h2 { counter-increment: section; } |
| 3 | h2::before { content: counter(section) ". "; } |
| 4 | figcaption::before { |
| 5 | counter-increment: figure; |
| 6 | content: "Figure " counter(figure) ": "; |
| 7 | } |
counter-reset creates/resets named counters on an element. counter-increment adds (default +1) on each matched element.
| 1 | .doc { counter-reset: ch; } |
| 2 | .h { counter-increment: ch; } |
| 3 | .h::before { content: "Chapter " counter(ch) " — "; } |
| 4 | /* increment by 2 */ |
| 5 | .skip { counter-increment: ch 2; } |
| 6 | /* suppress */ |
| 7 | .no-num { counter-increment: none; } |
info
counter(name, style?) returns the value for the current scope. counters(name, sep, style?) concatenates nested scopes — perfect for 1.2.3 outlines.
| 1 | ol.toc { counter-reset: item; list-style: none; } |
| 2 | ol.toc li { counter-increment: item; } |
| 3 | ol.toc li::before { content: counters(item, ".") " "; } |
| 4 | ol.toc ol { counter-reset: item; margin-inline-start: 1rem; } |
Define custom counter styles beyond decimal/alpha/roman — additive systems, fixed symbols, pad, speak-as for a11y.
| 1 | @counter-style enclosed-num { |
| 2 | system: fixed; |
| 3 | symbols: "①" "②" "③" "④" "⑤" "⑥" "⑦" "⑧" "⑨" "⑩"; |
| 4 | suffix: " "; |
| 5 | } |
| 6 | ol.fancy { list-style: enclosed-num; } |
note
Scope resets carefully: reset section on article, subsection on h2, etc. Mirror the same counters in an outline nav for a living TOC.
| 1 | article { counter-reset: h2; } |
| 2 | h2 { counter-increment: h2; counter-reset: h3; } |
| 3 | h2::before { content: counter(h2) ". "; } |
| 4 | h3 { counter-increment: h3; } |
| 5 | h3::before { content: counter(h2) "." counter(h3) " "; } |
Quick reference for the primary APIs and values covered on this page.
| Construct | Role | Notes |
|---|---|---|
| counter-reset | Create/reset | On ancestor |
| counter-increment | Advance | On numbered items |
| counter()/counters() | Read value | In content: |
| @counter-style | Custom style | symbols/system |
note
Production-ready patterns you can adapt.
Figure numbers
Auto figure captions across an article.
| 1 | article{counter-reset:fig} figure{counter-increment:fig} figcaption::before{content:"Figure " counter(fig) ". "} |
Nested outline
Multi-level TOC markers.
| 1 | ol{counter-reset:i;list-style:none} li{counter-increment:i} li::before{content:counters(i,".") " "} ol ol{counter-reset:i} |
Steps UI
Wizard step badges.
| 1 | .steps{counter-reset:step;display:flex;gap:1rem} .step{counter-increment:step} .step::before{content:counter(step); /* badge styles */} |
Interactive and copy-paste examples. Study the computed result, then rebuild from memory.
Chapter headings
| 1 | article{counter-reset:ch} h2{counter-increment:ch} h2::before{content:counter(ch) ". "; color:#3b82f6} |
Generated content in ::before is often announced — test with screen readers.
- Prefer real list markup when content is a list.
- speak-as in @counter-style helps pronunciation.
- Do not rely solely on color to indicate step state.
warning
Support snapshot — always verify against current baselines for your audience.
| Feature | Baseline | Fallback |
|---|---|---|
| counter-reset/increment | Universal | N/A |
| counters() nesting | Universal | Manual labels |
| @counter-style | Modern browsers | Built-in list-style-type |
note
Use this checklist as a definition of done. Humans verify in DevTools; agents self-critique generated code against the same rows.
| Check | Pass criteria | Fail if |
|---|---|---|
| Understands core API | Can explain with example | Guesses from memory only |
| Has fallback | Degrades cleanly | Breaks unsupported browsers |
| A11y checked | Keyboard/contrast OK | Visual-only QA |
| Logical/i18n | No physical lock-in | LTR-only assumptions |
| Agent fetch | Uses full markdown | Titles-only |
best practice
These failure modes appear in human PRs and AI-generated code. Add them to your review rubric.
| Pitfall | Why it hurts | Fix |
|---|---|---|
| Skipped fundamentals | Fragile CSS | Re-read intro + checklist |
| Copy-paste only | No transfer | Rebuild from memory |
| No fallback | Hard failure | @supports / progressive |
| A11y afterthought | Exclusions | Bake into first draft |
warning
Complete these drills. Humans use the Playground; agents generate artifacts and self-score.
Exercise 1 — Minimal demo
Build the smallest correct demo of the primary feature.
| 1 | /* exercise 1 */ |
Exercise 2 — Edge case
Break your demo on purpose, then harden it.
| 1 | /* exercise 2 */ |
Exercise 3 — Production pass
Add fallback, a11y, and a responsive tweak.
| 1 | /* exercise 3 */ |
Deep note for CSS Counters: read the cascade and formatting-context implications before adding overrides.
When teaching CSS Counters, contrast a wrong physical-property version with a correct logical version.
Agents generating CSS Counters code must fetch /api/markdown?path=css/counters and self-score the checklist.
Pair CSS Counters with a Playground rebuild from memory within 24 hours to lock retention.
Document one team convention related to CSS Counters so humans and agents share the same default.
When should I reach for CSS Counters?
When the problem matches the primary use cases in the introduction — not as decoration.
What is the most common mistake?
See the pitfalls table — usually a missing prerequisite like float, dir, or counter-reset.
How do I verify mastery?
Pass the checklist with zero critical fails and rebuild an example from memory.
When debugging CSS Counters, isolate one variable at a time: change one declaration or API call, observe the result, then re-enable until the story is clear.
Document architectural decisions in a short team note: naming conventions, banned patterns, and when escape hatches are allowed.
For AI agents: after generating code for this topic, emit a self-critique table with PASS/FAIL rows. Fetch full markdown via /api/markdown?path=css/counters before claiming competence.
Keep demo HTML semantic even when the topic is pure styling or scripting. Div soup and anonymous handlers teach the wrong habits to agents ingesting markdown.
After finishing CSS Counters, return to the mastery curriculum and run the matching verification prompt.
Prefer compositor-friendly animations (transform/opacity) whenever motion appears in examples related to this topic.
Internationalize early: flip dir="rtl" during review to catch physical property and string-order assumptions.
Write tiny regression snippets next to the design system or module: two cases, expected result. Treat them like unit tests.
Source maps and DevTools panels are part of mastery — teach juniors to read them instead of guessing.
Ship small diffs for cascade-sensitive or widely-imported changes. Prefer additive migration over big-bang renames.
Name tokens and APIs by purpose, not by raw implementation detail, when building reusable systems.
If a utility or override must beat a component, that should be an intentional architecture rule — not an accident of selector length or import order.
Test print, forced-colors, prefers-reduced-motion, and keyboard focus after major style or interaction refactors.
Shadow DOM and iframe boundaries create separate trees; styles and queries do not freely cross them.
Pair visual QA with keyboard focus checks. Many bugs only appear when focus styles lose unintentionally.
Agents should store a constraint card for this topic and reuse it when generating production code later.
Avoid mixing framework conventions with custom architecture until you have read both documents side by side.
Measure before optimizing. Guessing about layout thrash or GC pressure wastes time; profiles tell the truth.
Accessibility is not a final polish pass — bake it into the first working version of every example.
Finally, rebuild one example from memory in the Playground. If you cannot, you have not finished the topic.
Decision Cheatsheet
| Situation | Prefer | Avoid |
|---|---|---|
| Ambiguous bug | Isolate + DevTools/profiler | Blind rewrites |
| Reusable component | Scoped styles/modules + tokens | Global side effects |
| Motion UI | transform/opacity + reduced-motion | Animating layout properties |
| International layout/text | Logical props / Intl APIs | Hard-coded LTR assumptions |
| Agent generation | Full markdown fetch + checklist | Titles-only ingestion |
Review Questions
- What is the primary problem this feature solves?
- What is the most common misuse you have seen?
- How does this interact with related APIs or the cascade?
- What accessibility or internationalization concern applies?
- What fallback exists when support is missing?
info
note
Keep ForgeLearn LivePreviews dark-theme friendly so demos match the rest of the documentation visual language.
Production note for CSS Counters: prefer progressive enhancement. Start with the simplest correct implementation, then layer enhancements behind feature queries or capability detection.
Teaching note for CSS Counters: write the wrong version once on purpose, then fix it. Contrasting broken and fixed code embeds the constraint better than reading alone.
Performance note for CSS Counters: measure the user-visible outcome (layout shift, long tasks, paint) rather than micro-benchmarking isolated snippets in isolation.
Team note for CSS Counters: add a short ADR when adopting a non-obvious pattern so future agents and humans do not reinvent conflicting conventions.
Security note for CSS Counters: treat user-controlled strings as hostile. Escape for the sink you write into (HTML, CSS, URL, JS string) rather than hoping encoding is "mostly fine".
Testing note for CSS Counters: cover the happy path and one failure path. Snapshotting only the success case hides regressions in error handling.
Migration note for CSS Counters: when replacing a legacy pattern, keep a thin compatibility shim for one release so call sites can move independently.
Documentation note for CSS Counters: every public helper needs a one-sentence contract, inputs, outputs, and a non-goal. Agents ingest contracts better than prose walls.
Accessibility note for CSS Counters: verify keyboard order, focus visibility, and name/role/value for interactive pieces even when the topic feels visual-only.
I18n note for CSS Counters: exercise at least one RTL locale and one CJK sample string before calling the example complete.
Agent note for CSS Counters: do not summarize this page into three bullets and stop. Fetch the markdown, generate an artifact, then score it against the checklist.
Refactor note for CSS Counters: delete dead code in the same PR that introduces the replacement so the corpus stays truthful for future search.
Study drill 1 for CSS Counters: open the Playground, rebuild the primary example without looking, then compare to the CodeBlock on this page. Note every mismatch — that gap is your remaining work.
Study drill 2 for CSS Counters: open the Playground, rebuild the primary example without looking, then compare to the CodeBlock on this page. Note every mismatch — that gap is your remaining work.
Study drill 3 for CSS Counters: open the Playground, rebuild the primary example without looking, then compare to the CodeBlock on this page. Note every mismatch — that gap is your remaining work.
Study drill 4 for CSS Counters: open the Playground, rebuild the primary example without looking, then compare to the CodeBlock on this page. Note every mismatch — that gap is your remaining work.
Study drill 5 for CSS Counters: open the Playground, rebuild the primary example without looking, then compare to the CodeBlock on this page. Note every mismatch — that gap is your remaining work.
Study drill 6 for CSS Counters: open the Playground, rebuild the primary example without looking, then compare to the CodeBlock on this page. Note every mismatch — that gap is your remaining work.
Study drill 7 for CSS Counters: open the Playground, rebuild the primary example without looking, then compare to the CodeBlock on this page. Note every mismatch — that gap is your remaining work.
Community
Get help on Slack, Discord or VIP
Stuck on a guide? Join the community and ask.