Currently every feature description has to appear in both the readme and its file:

 - [](# "pr-jump-to-first-non-viewed-file") [Jumps to first non-viewed file in a pull request when clicking on the progress bar.](https://user-images.githubusercontent.com/16872793/85226580-3bf3d500-b3a6-11ea-8494-3d9b6280d033.gif)

void features.add({
	id: __filebasename,
	description: 'Jumps to first non-viewed file in a pull request when clicking on the progress bar.',
	screenshot: 'https://user-images.githubusercontent.com/16872793/85226580-3bf3d500-b3a6-11ea-8494-3d9b6280d033.gif'
}, …);

We could instead just parse it from the readme. To parse the data, I think we just need to change this function: Instead of loading each tsx file, we lookup the supplied id in readme.md and parse that

https://github.com/sindresorhus/refined-github/blob/42dfa16cd5366de26b475de94f8225003df07364/webpack.config.ts#L14-L37

Advantages

• no need to maintain duplicate content
• the build will effectively enforce documentation, or else it will fail (in watch mode this should probably just log an error)

Challenges (easy)

• some features appear in the highlights table, so both need to be parsed
• some features have 2 links in the readme, so either we pick one or allow multiple links in the options
• we need to find another way to disable features. I think we can hardcode this list in webpack.config.ts so it can be injected exactly like it is now

Resulting code

 - [](# "pr-jump-to-first-non-viewed-file") [Jumps to first non-viewed file in a pull request when clicking on the progress bar.](https://user-images.githubusercontent.com/16872793/85226580-3bf3d500-b3a6-11ea-8494-3d9b6280d033.gif)

void features.add(__filebasename, …);