What the agent does
The agent reads the manifest and the four entry points first. In an extension like chibat/chrome-extension-typescript-starter, public/manifest.json is Manifest V3 with an action whose default_popup is popup.html, an options_ui page, a content_scripts entry matching <all_urls> that loads js/vendor.js and js/content_script.js, a background.service_worker at js/background.js, the storage permission and <all_urls> host permission. webpack/webpack.common.js builds popup, options, background and content_script from src/ into dist/js, splits a vendor chunk for every entry except background, and copies public/ over dist/. src/popup.tsx is a React 18 component that writes a counter to the badge with chrome.action.setBadgeText and sends { color: "#555555" } to the active tab with chrome.tabs.sendMessage; src/content_script.tsx receives it in chrome.runtime.onMessage, sets document.body.style.backgroundColor and answers with sendResponse. src/background.ts only schedules an empty polling every 30 seconds, and src/options.tsx reads and writes favoriteColor and likesColor in chrome.storage.sync.
Shortcuts live in the manifest under commands. The agent adds the reserved _execute_action command so a key opens the popup, and a named toggle-background command with a suggested_key of Ctrl+Shift+Y by default and Command+Shift+Y on macOS, each with a description Chrome shows on its shortcuts page. Chrome lets an extension suggest at most a few defaults, and a suggested key that another extension already holds is left unassigned, so the description text matters and the options page needs to show what is bound.
The handler goes in the service worker. src/background.ts gets a chrome.commands.onCommand.addListener that, for toggle-background, queries the active tab in the current window and sends the same message the popup sends. To keep the popup and the worker in step, the agent moves the message shape into a new src/messages.ts with a colorMessage(color) builder and a nextColor(current) helper that alternates between the tint and the page's own colour, and both popup.tsx and background.ts import it. The content script answers a reset by clearing the inline style rather than leaving the page grey. src/options.tsx gains a section that calls chrome.commands.getAll and lists each command with its current shortcut or an unassigned note, and a link to chrome://extensions/shortcuts, which an extension page can open only through chrome.tabs.create.
Verification follows .github/workflows/build.yml, which runs npm ci, npm run build and npm test on Node 20 and 22. npm run build runs webpack in production mode and fails on a type error, and @types/chrome at 0.0.158 already declares chrome.commands. npm test runs Jest through ts-jest with tsconfig.test.json over src/, where src/__tests__/sum.ts is the only test, so the agent adds src/__tests__/messages.ts for nextColor and colorMessage, which need no browser, and src/__tests__/background.ts with a stubbed chrome global whose commands.onCommand.addListener captures the listener so the test can fire toggle-background and assert on tabs.sendMessage. It then loads dist/ as an unpacked extension in the sandbox's Chromium, presses the shortcut on a page and reads the result.
Start it with one API call
POST https://api.phoenix.so/o/{org}/sessions with this body:
{
"agent": "claude",
"input": "A Linear issue labelled agent says: \"Add keyboard shortcuts. One to open the popup and one to toggle the page background, and show the current bindings on the options page.\" Do the work in this repository. Read public/manifest.json, src/background.ts, src/popup.tsx, src/content_script.tsx, src/options.tsx, webpack/webpack.common.js, jest.config.js, tsconfig.test.json, src/__tests__/sum.ts and .github/workflows/build.yml first. Add a commands section to the manifest with _execute_action and a toggle-background command with suggested keys and descriptions. Handle toggle-background in the service worker with chrome.commands.onCommand, sharing the message shape with the popup through a new src/messages.ts, and make the content script able to clear the colour again. On the options page, list each command with its current shortcut from chrome.commands.getAll and a link that opens chrome://extensions/shortcuts in a new tab. Add Jest tests for the message helpers and for the command listener with a stubbed chrome global. Run npm ci, npm run build and npm test, load dist as an unpacked extension in a browser and try the shortcut, and open a pull request that lists the commands and their default keys.",
"resources": [
{
"url": "https://github.com/your-org/your-app"
}
]
}- Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
- Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
- Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.
What comes back
What comes back is a pull request that changes public/manifest.json, src/background.ts, src/popup.tsx, src/content_script.tsx and src/options.tsx, and adds src/messages.ts with two test files under src/__tests__/. The description lists each command, its description string and its suggested keys per platform, and quotes the build and Jest summaries.
The transcript holds the npm ci, the webpack build with the chunk list, the Jest run, and the browser session where the agent loaded dist/ unpacked and pressed the shortcut, with a screenshot of the page before and after, so a reviewer sees the command reach the content script rather than reading that it should. The repository's build workflow runs the same install, build and tests on the pull request.
Before merging, a reviewer checks that the manifest's suggested keys do not collide with Chrome's own bindings, that the listener ignores a command it does not know, that the options page copes with an unassigned shortcut, and that the content script restores the original background rather than a hard-coded colour. A follow-up message to the same session adds a third command or a per-site toggle stored in chrome.storage.sync, and the sandbox already has node_modules and the browser.
Run it on every Linear issue labelled agent
In the Project that holds the extension, add an automation with this prompt and a linear trigger on linear.issue.labeled filtered to the agent label. Each labelled extension issue starts its own run in its own sandbox and ends in its own pull request; the issue text follows the prompt as data, and the trigger fills in no placeholder.
A handler on your own server does the same without the automation: it receives Linear's webhook, puts the issue text into input, and sends POST /o/{org}/sessions with the repository in resources and a request_id that is a UUID v5 of the issue id, so a redelivered webhook returns the run already started.
The use case this job follows, with its automation and trigger: Turn a Linear label into a PR.
Why run it on Rush?
Extension work is checked by loading dist/ unpacked, reloading after each build and pressing keys in a real page, and the service worker's logs live in a separate inspector. In a Rush sandbox the agent has Node, the webpack build and a Chromium it can drive, loads the extension itself, presses the shortcut and keeps the screenshots and the worker's output in the record.
Small extension requests pile up because each one costs a build-and-reload cycle on someone's machine. Labelling the issue hands that cycle to a sandbox, and the pull request arrives with the build, the tests and a browser check already done. Several extensions get the same treatment at once, each in its own sandbox.
Frequently asked questions
Can users change the keys?
Yes. Chrome lets users rebind any extension command at chrome://extensions/shortcuts, and the options page the agent writes links there and shows the current binding from chrome.commands.getAll, so a user sees what is bound after changing it.
What happens if my suggested key is already taken?
Chrome leaves the command unassigned rather than overriding another extension or its own binding. The options page shows it as unassigned with the link to bind it. The agent picks defaults that Chrome's documentation does not reserve, and the pull request lists them so a reviewer can change them.
Does the shortcut work when the popup is closed?
Yes. chrome.commands.onCommand fires in the service worker, which Chrome wakes for the event, so toggle-background works with no popup open. _execute_action is handled by Chrome itself and opens the popup.
How is the service worker tested without Chrome?
The command listener is a function registered on chrome.commands.onCommand, so the Jest test installs a stubbed chrome global, captures the listener and calls it with toggle-background, then asserts that chrome.tabs.sendMessage was called with the shared message. The browser check in the sandbox covers the real wiring.
More Chrome Extension jobs
The same job on other stacks
More jobs like this: Turn a Linear label into a PR
- Add KV Caching to a Cloudflare Worker (Cloudflare Workers)
- Add Lambda Powertools to an AWS CDK App from a Linear Issue (AWS CDK)