Phoenix Code's npm test prints a message instead of running tests
The text editor designed to make coding as simple and fun as playing a video game
At a glance
- What is it?
- Phoenix Code is a browser-based text editor forked from Adobe's Brackets, served from a static web server with no compile step, and its file is a build and testing guide rather than a description. The sharpest detail is in the manifest: the default test script does not run tests, it prints a message pointing you back at the readme. Around that sit four documented ways to run the suite, a directory named as if it were a file, and two source trees plus a markdown viewer.
- Who is it for?
- Phoenix Code fits a team that wants a web editor with a Brackets-shaped extension model and no toolchain at all on the developer's machine, which is the same reason the editor can be served from a static file server. Two things to check before you adopt it.
- Can I use it commercially?
- Yes, with strict conditions. AGPL-3.0 is a network copyleft licence: if people use a modified version over a network, for example as a hosted service, you must offer them its source code under the same licence.
- Is it still maintained?
- Yes. The repository last received commits 13 days ago.
- What is it written in?
- Mainly JavaScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- phcode-dev/phoenix3,024 ★ · +32
| Repository | 2026-09-24 | 2026-11-15 |
|---|---|---|
| phcode-dev/phoenix | 3k | 3k |
32 Stars in the last 52 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
Three names and a fourth domain
The project is called Phoenix Code, the repository is named after a single word, the website is on one domain and the manifest's homepage field points at a different one entirely.
The release titles say Phoenix Code. The description says the editor is designed to make coding feel like playing a video game. The website in the readme is a short domain matching the repository owner. And the manifest's homepage field is a completely different company domain, with the repository block beside it carrying an empty branch string and an empty commit identifier.
That combination is more than untidy naming, because search engines and registries both read those fields. Someone looking for the editor by name finds the right repository; someone reading the package metadata finds a homepage for a different product.
The manifest also carries two version fields. The package version is a three-part number with a trailing dash and a zero, and beside it is a separate API version with the same three parts and no suffix. Nothing in the visible file explains what the suffix means or which one a consumer should compare against, and the release list uses a fourth numbering style with two components and a stable-release label.
So there are four version representations in one project and no note reconciling them.
The tenets list has two items numbered two
The principles section is nine items long and is numbered one to eight.
There are two items carrying the number two. Everything after that shifts by one, so the list ends at eight when it has nine entries.
Reading the list as content rather than as formatting, it is a good description of what makes this editor different from a desktop one. It targets web development, with three file types given special status. It treats the editing experience as a game. It is a light-weight editor. It keeps compatibility with the Brackets extension model, excluding the extensions that need a local Node runtime. It insists on an uncompromised local development experience. It supports pluggable remote back-ends. Its core works from a static web server. Most code changes need no recompilation. And it prioritises simplicity.
Those are nine commitments and they are the actual specification of the project. Everything else in the file is how to build it and how to test it.
The one that explains the rest is the static server plus the compile step point. An editor whose core is a set of files that a browser can load is an editor whose development loop is a file watcher and a reload, and that is why the testing section has four sections rather than one.
The default test command prints a message
Look at the default test script in the manifest and it does not run tests.
It prints a message telling you to see the section of the readme titled Running and debugging tests. That is the whole script.
So `npm test`, the command every continuous integration system and every developer reaches for by reflex, does nothing and exits successfully. Nothing in the manifest makes it fail, so a pipeline that runs it as a smoke check would go green without having executed a single assertion.
The other scripts around it are ordinary. Lint runs on two directories with warnings suppressed. A fix variant rewrites in place. One script zips the test files through the build tool. One updates pinned versions for the language servers. One starts a static file server for test files on a fixed port with caching disabled.
The post-install hook is the other thing in that block worth noting. It changes into a nested directory and runs an install there, so installing the editor also installs a second package inside it, which is the automation integration the readme points at for two assistant clients.
Four ways to run the suite, and the file recommends the slowest
The testing section is the longest part of the readme, and it documents four routes.
The first is in the browser, and the file calls it the easiest and preferred way. You build, navigate to a page, and open a test runner from a menu path inside the editor's own diagnostics menu, with a menu item to reset the test data files and the browser's own developer tools for debugging.
The second is headless, through Playwright, in continuous integration. The file is explicit that Playwright is not the test framework: it is used as a headless browser, two of them, to run tests written in two other frameworks. There is a command per browser, a debug variant per browser, and an environment variable that selects one of four named integration suites, three of which are named after internal views rather than after behaviour.
The third is the same thing against a release build rather than a development one, with a third command suffix.
And then the recommendation, which is the part to notice. The file says that to actually fix a failing test you should use the first route, that the browser route has much better debugging ergonomics, and that fixing it in the browser will almost certainly fix it in the headless runner. So the project's stated workflow is to reproduce a failure in the slowest environment on purpose, in exchange for being able to see it.
That is a defensible trade and it is also why the default script refuses to pick one for you.
A directory whose name ends in a file extension
The top level of the tree has a directory named with a file extension on it: the build definition lives in a directory whose name ends in the extension used for the file that would normally hold it.
Everything else in the listing is ordinary, which makes it stand out. There is a source directory, two more source directories beside it, a build directory, a docs directory, a test directory, a separate directory for the headless browser tests, and one for the automation integration.
Three of those source directories are worth separating in your head. One holds the editor itself, one holds the parts that need a local Node runtime, and one holds a markdown viewer. That split is the practical shape of the project: a browser editor with a small Node side, plus a viewer that is its own thing.
Two files at the root are also unusual in a different way. One is a configuration for the headless test runner. The other is a JavaScript file whose name ends in a configuration suffix and that sits next to a proxy server script, which together are how the development server forwards requests.
There is a small JSON file for tracking other repositories, which is the sort of file that exists to answer a question about where a bug should be filed rather than anything about the editor.
No runtime dependencies, an embedded database, and a hook on install
The manifest has no dependencies section at all. Every entry is a development dependency.
That is the packaging consequence of the compile step point. The editor is not bundled from a dependency tree at install time; it is source that runs from a file server, so the packages listed are the tools used to build, lint, test and package it rather than libraries the editor loads at runtime.
Among them are twelve plugins for the build tool, the two browser drivers for the headless tests, two testing frameworks and a reporter, the commit message linter and its configuration, the hooks runner, a static file server, a proxy library, a documentation-to-markdown generator, and an embedded key-value store.
That last one is the interesting entry. An embedded key-value store in a browser editor is what backs the virtual file system you can browse from a debug menu inside the editor. So the editor's in-memory and on-disk file tree is a database rather than a directory of files, which is what lets the whole thing run from a static server with no server-side filesystem.
Two build-tool details are easy to miss. The clean script is described as cleaning builds only, and the release scripts come in three named variants, one per deployment stage, which is why the testing section can point a headless run at a built stack rather than at a development server.
A Brackets fork, with a typo in the copyright line
The acknowledgements section is four credits and they describe a stack rather than a project.
The editor is based on the Brackets code editor by Adobe. Its main code editing library is a separate editor component. It was inspired by an earlier project whose stated purpose was to port Brackets to the web. And the server side in the browser is based on yet another small server project.
So the ancestry runs through a discontinued desktop editor to a web port of it, and this project is the second generation of that lineage. That matters for extension compatibility, which the principles list states as full compatibility with the Brackets extension model minus the extensions that need a local Node runtime, which is exactly the set of extensions that could not survive being ported to a browser.
The licence section then has a typo in its second copyright line, where the base project is named with its vowels transposed. It is a small thing in the one place a reader looks to confirm provenance.
The copyright holder named there is a company that is not the repository owner, the website, or the domain in the manifest's homepage field. So the file that establishes who wrote this code points at a fourth identity, alongside the one in the copyright year range that begins in 2021.
Editorial conclusion
Phoenix Code fits a team that wants a web editor with a Brackets-shaped extension model and no toolchain at all on the developer's machine, which is the same reason the editor can be served from a static file server. Two things to check before you adopt it. Plan your test loop first, because the project documents four different ways to run its suite and explicitly recommends the slowest one for debugging, which is a choice about developer time rather than about correctness. And read the version fields in the manifest before you automate anything about it, because the package carries both a version with a trailing suffix and a separate API version, and the repository's own homepage points at a domain unrelated to the editor.
Frequently asked questions
what is Phoenix famous for?
That question is about the city, the mythological bird, or a game character, none of which this repository is. This project is a browser-based text editor derived from Adobe's Brackets, designed for web development, served from a static web server, and built to run extension packages written for that earlier editor.
how to install phoenix
There is no installation package. The build instructions are to install a build tool's command line interface globally, with a privileged variant on Mac and Linux and an unprivileged one on Windows, then run the dependency install in the repository and a build script. Running it means starting a local server and opening a path in a Chrome or Edge browser.
how to use phoenix
As a web page. The readme's running instructions are to start the local server script and navigate to a source path in Chrome or Edge, with a separate path for a test runner page. There is no desktop application to install and no server-side component to run, because the core is designed to work from a static web server.
Is Phoenix a Mexican city?
No, and the repository is not about that city either. This is a text editor project whose releases are titled with a two-word product name, whose package manifest carries both a package version and a separate API version, and whose website, repository owner and manifest homepage are three different names.
What is the meaning of a Phoenix?
The repository does not discuss it. Its own description is a text editor designed to make coding feel like playing a video game, and its principles list that comparison as its second tenet.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/phcode-dev-phoenix)Star history chart for your README
[](https://hysenlabs.com/projects/phcode-dev-phoenix)Was this useful?