AI Services Market › Use cases › Software & DevOps
Software & DevOpsDone with Claude Code 2026-10-20
Let Claude Code run a Chrome extension project
The author gave Claude Code the CEO role for a free Chrome extension. In week one it had him rename it and built the next version, but it still needs his approval before each post.
- What They Did
- The author set Claude Code to run nightly at 9 pm in the desktop app with a self-edited markdown playbook. Using Playwright on a signed-in Chrome profile, it checked the Chrome Web Store dashboard, comments, and subreddits, planned posts, worked on the extension, and wrote a pinned Notion report; the human still had to reply "go" before anything public was posted.
- How It Went
- In its first week it pushed a rename after spotting a near-duplicate featured extension, fixed stale privacy copy, and built a new version with onboarding and a window-stashing feature that was sent for store review after 67 unit and 55 end-to-end checks. Results were still tiny: 4 weekly users, 14 installs in 30 days, 0 reviews, $0 spent, and a release sat five days when Google forced a re-login.
- Worth Knowing
- The main catch was authentication: Google re-prompted for the store password every few days, and the author later added upfront checks so failed logins show up at the top of the nightly report.
Try it yourself with Claude Code
Act as the CEO of my project [project name], a free [type of tool], for the next week. Look at the current state in [folder or repo] and propose three priorities for the week with reasons. Then build the first one in the code and write a short update on what changed. Anything public, like a post, a release, or a rename, needs my approval first, so show me the draft and wait.
Do this with a skill.
These AI Services Market playbooks do this exact job.