Scheduled web scraping.
Watching a competitor's prices by hand is a chore you'll do for three days and then quietly abandon. The SaaS tools that promise to do it for you charge per tracked URL and lock the history behind their export tier. This blueprint hands you the scraper instead. You describe the pages and the cadence, and the agent builds a task that checks prices on a schedule and keeps the record in a file you own.
A scheduled scraping task built on Puppeteer that visits your list of competitor pages every six hours, reads the current product price off each one, and compares it to what it saw last time. When a price moves, it appends the change to a JSON log, so you end up with a running history of who moved what and when, not just a snapshot of right now.
You watch it happen in the task workspace: the agent scaffolds the project, writes the scraping logic, runs it once against the target pages to confirm it pulls real numbers, and shows you the log output. The code lands on a branch in your repository with a pull request ready.
The first version tracks price and cadence and is yours to steer with plain follow-up messages:
Each follow-up wakes the same repository and the agent continues on its own code.
Do I need to know how to code? No. You describe what to watch and how often; the agent writes the scraper, runs it once to prove it works, and ships the code to your repository.
Where does the price history live? In a JSON log in your own repository. The data is yours, with no per-URL fee and no export paywall.
What if a site changes its layout? The selectors live in your code. Tell the agent the page broke and it updates the scraping logic, same repo, next run.