NullToHero

See what it builds

nulltohero.netlify.app is the plugin's own site, and it is the honest answer to "what can I actually get out of this". Nothing on it is a mockup: it is a scroll-driven story that writes the logo onto a blank sheet, opens a terminal, raises a generic template page, then audits that page on screen and corrects it.

_{Open the live version: the motion is scroll-driven, so it answers your wheel rather than a timer, and the capture above only walks it at one fixed pace}

The site holds itself to the rules it ships. Its source carries the plugin's own audit annotations: where a low-contrast element is deliberate, it declares data-contrast-exempt-reason and says why, which is the same mechanism /audit reads when it decides whether a contrast finding is a defect or a documented exception. The one place it shows unreadable text on purpose is the specimen page being audited, and that is written down in the code rather than left for a reviewer to guess.

You can check the claim rather than take it:

```bash node null-to-hero/tools/audit/gate.mjs https://nulltohero.netlify.app/ --min-score 90

Install

NullToHero is a Claude Code plugin and a marketplace in one repository. The marketplace manifest sits at the repository root; the plugin itself lives in the null-to-hero/ folder.

A. From the marketplace (recommended, auto-updates)

/plugin marketplace add MariusYvard/NullToHero
/plugin install null-to-hero@null-to-hero-marketplace

 Design and build

Your design partner. It plans the look, builds the pages, fixes spacing and type, makes everything responsive, and adds tasteful motion. You describe the goal, it produces real, production-ready front-end.

<details> <summary><b>All 35 commands</b></summary>

Common runs: a new page (setup → shape → build → layout → adapt → amplify → harden), a refresh (improve → polish), after an audit (fix → ship), a design system (document → extract → tokens).

</details>

then: Kimi Work → Plugins → Personal (「个人」) tab → ＋ on NullToHero

```

The agents package is the answer to "what about Cursor". Those hosts differ only in the directory they read, and they all read the shared one:

Amp, Goose, Roo Code, Factory Droid, Kiro, Junie, Letta, OpenHands and the rest of the standard's client list read the same file; check your host's page for the directory it prefers.

codex and agents land in the same directory, so they exclude one another and the installer refuses rather than overwriting. Pick codex if you want the fifteen sub-agents; pick agents for anything else.

The skills install as nth-seo, nth-siteasy, nth-audit and nth-cms, because a skills directory is shared with every other pack on the machine and audit is a name someone else will claim. The deterministic tools and the asset library are read from the clone, so keep it where it is.

Claude Code is unaffected by any of this. It keeps loading null-to-hero/ directly, the prose keeps naming Claude's tools, and the build substitutes them only in the generated packages. tests/portability.mjs fails if that stops being true.

Each package was installed and read by a running host, never by a specification alone: Codex 0.148.0 and Kimi Code 0.37.2 on 2026-08-19, opencode 1.18.21 and Gemini CLI 0.56.0 on 2026-08-23 and 24. dist/VERIFY.md records what those runs showed, including the part that disappoints: both standard hosts loaded two deliberately invalid skills without complaint, one whose declared name does not match its folder and one with a 1024-character limit overrun. So a host accepting the package proves it is discovered and says nothing about whether it conforms; that is held by tests/portability.mjs instead. bash tests/verify-hosts.sh reproduces the first two runs.

How NullToHero compares

NullToHero overlaps four kinds of tool. The honest comparison: each column is excellent at what it is for, none of them spans the whole loop of building, auditing, scoring, fixing and handing over inside your own files.

<p align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="docs/compare-dark.svg"> <img src="docs/compare-light.svg" width="920" alt="Comparison matrix: NullToHero vs AI site builders (v0, Lovable, Bolt), audit tools (Lighthouse, axe), SEO suites (Semrush, Ahrefs) and design skills (ui-ux-pro-max, ux-ui-mastery, Taste) across 13 capabilities, including handing the finished site to a non-technical owner. NullToHero is the only column with every capability checked."> </picture> </p>

<details> <summary><b>Text version of the table</b></summary>

_{✅ yes · 🟡 partial · — no. The figure above says the same thing with a tick, a half disc and a dash, so it survives greyscale. Both are drawn from null-to-hero/tools/data/compare-matrix.csv. Nuances: Lighthouse's deterministic score covers performance, not design or content; Semrush's Site Health score is deterministic but proprietary and not CI-native; builders generate tasteful UI without judging or scoring it.}

</details>

NullToHero is the one that spans build, defects, SEO, a scored whole-site audit and the handover in a single plugin, with every finding routed to the command that fixes it. It is not a hosted product or a visual editor: it runs inside Claude and edits the real files in your project, so the output is yours to keep and version.

---