Goal: Detect meaningful price changes on three competitor pages each day and email me only when a real change occurs. Inputs available: A headless browser, email-send capability, the three page URLs (you'll need to provide them), and a stored record of yesterday's prices (missing, so the workflow has to create and persist it). Workflow steps: 1. Load prices. Job: fetch each of the three pages and extract the price. Input: the three URLs. Output: today's price per page. Tool: headless browser. Single-purpose. 2. Load baseline. Job: read the last recorded price per page. Input: the stored price file or row. Output: yesterday's price per page. Tool: a small persistent store (file or key-value). Flag: this store must exist; on first run there is no baseline, so treat run one as baseline-only. 3. Compare. Job: compute the change per page and decide if it clears the threshold. Input: today's and yesterday's prices. Output: a list of pages with a real change. No external tool. 4. Notify. Job: if any page cleared the threshold, send one email summarizing the changes. Input: the changed-pages list. Output: a sent email. Tool: email. 5. Persist. Job: write today's prices as the new baseline. Input: today's prices. Output: updated store. Tool: the store. Handoffs: Step 1 and 2 outputs both feed step 3; step 3's output gates steps 4 and 5. Extraction in step 1 should run page-by-page in isolation so one broken page doesn't fail the others. Failure handling: Step 1 (browser) is the likely failure, from a page layout change or a load timeout. On extraction failure for a page, skip that page, log the reason, and include "could not read page X today" in the email rather than failing silently. Email-send failure: retry once, then write the alert to the log so it isn't lost. Verification: Confirm the email fired only when at least one page cleared the threshold, and that a no-change day sends nothing. Spot-check that the persisted baseline equals today's extracted prices. What would make this simpler: Define "real price move" as a fixed threshold (for example, a change of more than $1 or more than 1%). With a clear threshold this is a single daily script, not a multi-agent system, one call that fetches, compares, and emails.