This project starts from three public contributions that deserve clear credit:
visual-explainerby Nico Bailon, credited as the upstream HTML artifact capability this repo can install and build around.- "The unreasonable effectiveness of HTML" by Thariq S. Bate, credited as the conceptual foundation for using HTML when it keeps humans more in the loop than Markdown.
skillsby Matt Pocock, credited as inspiration for small, composable skill-shaped agent instructions.
html-explainer is an independent complementary layer. It does not replace those projects, claim affiliation, or copy their work as product output. It packages an intent-first workflow around them so Claude Code can decide when HTML is worth using, inspect evidence first, and end with a useful next prompt.
See CREDITS.md for the full attribution.
html-explainer helps Claude Code create HTML artifacts that people actually read.
It is not "HTML because it looks nice." It is HTML when a browser view helps the user understand, decide, review, compare, tune, or continue work better than Markdown.
Core path:
intent -> evidence -> visual understanding -> decision -> expert next prompt
visual-explainer gives Claude the HTML artifact capability.
html-explainer adds the operating discipline: when to use HTML, what evidence to inspect, how to keep the user in the loop, and how to end with the next useful prompt.
Open examples/before-after-decision.example.html.
It shows the same request two ways:
- a skimmable answer that looks organized but leaves the user to do the hard thinking;
- an intent-led HTML artifact that exposes evidence, risk, recommendation, and a copy-ready next prompt.
The point is simple: HTML is worth the extra tokens only when it removes mental work.
See docs/DEMO.md for the demo explanation.
First run a preview:
git clone https://github.com/ramsani/html-explainer.git /tmp/html-explainer
cd /tmp/html-explainer
DRY_RUN=1 bash install.shThen install:
bash install.shRestart Claude Code after installation.
The installer adds:
- the
thariq-html-effectivenessskill; - intent-first slash commands;
- 7 short core docs, advanced reference docs, patterns, and bundled examples under
~/.claude/html-explainer/; - a short managed guide in
~/.claude/CLAUDE.md; - backups under
~/.claude/html-explainer/backups/<timestamp>/.
The CLAUDE.md guide is marked with:
<!-- html-explainer:start -->
<!-- html-explainer:end -->
Future installs update only that block. Existing CLAUDE.md content is not overwritten.
If you already cloned the repo:
cd /tmp/html-explainer
git pull origin main
DRY_RUN=1 bash install.sh
bash install.shIf the old clone is gone:
rm -rf /tmp/html-explainer
git clone https://github.com/ramsani/html-explainer.git /tmp/html-explainer
cd /tmp/html-explainer
DRY_RUN=1 bash install.sh
bash install.sh/pick-the-right-html
/make-the-right-html
/check-the-plan
/check-the-diff
/reenter-project
/build-decision-tool
/audit-html
| Command | Use it when you want to... |
|---|---|
/pick-the-right-html |
Decide whether HTML is worth it and which artifact fits. |
/make-the-right-html |
Generate the right verified HTML artifact. |
/check-the-plan |
Review a plan before coding. |
/check-the-diff |
Review a diff or PR before accepting it. |
/reenter-project |
Understand a repo quickly and see the next action. |
/build-decision-tool |
Build a temporary editor, triage board, tuner, or config tool. |
/audit-html |
Check whether an existing artifact is actually useful. |
Start by choosing the right artifact:
/pick-the-right-html revisa esta tarea y dime qué artefacto HTML conviene crear. No generes todavía el HTML.
Then generate it:
/make-the-right-html genera el artefacto HTML correcto con evidencia, riesgos, incertidumbre, siguiente acción y prompt listo para copiar.
For common repo work:
/reenter-project ayúdame a entender este repo y ver sus áreas de oportunidad usando HTML.
/check-the-plan revisa este plan contra el repo real antes de implementar.
/check-the-diff revisa el diff actual y dime qué aceptar, corregir o revertir.
/build-decision-tool convierte esta decisión en una herramienta HTML editable con exportación.
Generated artifacts should be saved in a practical local path, opened in the browser when the environment allows it, and returned with a clickable absolute path.
Use HTML when the output replaces something people would skim or misunderstand:
| Work type | HTML should add |
|---|---|
| Exploration & Planning | Side-by-side options, trade-offs, dependencies, recommendation. |
| Code Review | Annotated diff, risk map, findings by severity, next action. |
| Design | Visual directions, component states, tokens, responsive states. |
| Prototyping | Clickable flows, sliders, knobs, live preview, export. |
| Diagrams | Architecture, workflows, failure paths, boundaries. |
| Decks | Shareable decision or progress story. |
| Research | Criteria, evidence quality, recommendation by use case. |
| Reports | Timeline, scorecard, risks, reentry checklist. |
| Custom Editors | Drag, sort, tune, validate, then copy the result out. |
Do not use HTML for short answers, single commands, tiny facts, or low-consequence notes.
Every artifact should:
- answer the user's main intent directly;
- cover obvious secondary needs when they affect the decision;
- show the evidence inspected;
- separate facts, inferences, assumptions, and unknowns;
- use the smallest useful artifact size;
- explain why HTML beats Markdown for this case;
- respect system light/dark mode;
- keep visual design minimal, flat, readable, and professional;
- end with a clear next action and an editable next prompt.
If interaction is included, it must change something meaningful and export usable output: Markdown, JSON, config, issue body, checklist, or prompt.
The default agent path is intentionally small:
DECISION_GATE -> PATTERN_GUIDE -> pattern file -> FACT_SHEET -> STYLE -> QUALITY_BAR -> CHAIN -> DELIVERY
Detailed older docs live in docs/reference/. They are available when needed, but they are not the default path.
visual-explainer is excellent at generating rich HTML artifacts.
html-explainer is for users who want more control over the reasoning before and after the artifact:
- intent selection before generation;
- evidence-first workflow;
- pattern routing;
- quality bar;
- lean artifact budget;
- next-step prompts that preserve context;
- safe installer and uninstaller;
- brief
CLAUDE.mdguide so the agent remembers how to use the system.
Use visual-explainer when you mainly want quick HTML output.
Use html-explainer when the HTML should help make or review a real decision.
- Not a standalone web app.
- Not an official Anthropic or Claude Code project.
- Not a replacement for
visual-explainer. - Not a general visual design tool.
- Not HTML for every answer.
Use it when HTML helps the user make or review a real decision.
Contributions are welcome when they strengthen the core path:
intent -> evidence -> visual understanding -> decision -> expert next prompt
See CONTRIBUTING.md before opening a PR.
Preview:
DRY_RUN=1 bash uninstall.shRestore latest backup, or remove managed files if no backup exists:
bash uninstall.shRemove without restoring backup:
RESTORE_BACKUP=0 bash uninstall.shRemove and delete backups:
KEEP_BACKUPS=0 bash uninstall.shThe uninstaller removes only the managed CLAUDE.md block. It does not rewrite the rest of your memory file.
# Skip upstream visual-explainer
INSTALL_UPSTREAM=0 bash install.sh
# Optionally download external Thariq reference pages
FETCH_EXAMPLES=1 bash install.sh
# Install into another Claude home
CLAUDE_HOME="$HOME/.claude" bash install.shbash -n install.sh uninstall.sh scripts/*.sh
scripts/validate-patterns.sh
scripts/validate-commands.sh
scripts/validate-examples.sh
DRY_RUN=1 INSTALL_UPSTREAM=0 FETCH_EXAMPLES=0 bash install.sh
scripts/smoke-install.sh
scripts/smoke-uninstall.shThis project is an independent integration layer inspired by and built around:
- visual-explainer by Nico Bailon;
- Thariq S. Bate's HTML effectiveness approach;
- Matt Pocock's skill and workflow style for practical engineering prompts.
It is not an official project of Nico Bailon, Thariq S. Bate, Matt Pocock, Anthropic, or Claude Code.
See CREDITS.md and CHANGELOG.md.