This is a submission for the Runner H "AI Agent Prompting" Challenge
What I Built
I built a semi-autonomous Wishlist Monitoring Agent powered by H Company's Runner H, designed with a "human-in-the-loop" element. This agent streamlines the tedious process of tracking desired products across various online retailers for price drops, stock availability, and the best available deals, while ensuring human oversight.
Problem Solved (with Human-in-the-Loop Context):
Time Consumption & Missed Opportunities: The agent automates the constant, repetitive checking of e-commerce websites, freeing up user time and ensuring optimal buying opportunities (price drops, stock alerts) are not missed.
Accuracy & Trust: For new products added by name, the agent intelligently finds and suggests the primary product URL. This critical "human-in-the-loop" step allows the user to quickly review and confirm the agent's initial identification, ensuring the correct item is being monitored before continuous automation begins.
Robust Error Handling: If the agent encounters an issue (e.g., a product page disappears, or it cannot extract data), it logs this clearly in the Google Sheet for human review, preventing silent failures and enabling quick manual intervention.
Solution Overview:
My Runner H agent acts as an intelligent personal shopping assistant. Users simply list desired products by name in a Google Sheet. The agent then autonomously:
For new products (added by name): It searches for and suggests the primary product page URL, which the user then reviews and implicitly confirms (or manually corrects) in the Google Sheet.
For confirmed products: It monitors the product's current price and stock status.
Searches for and compares prices of the same product across a curated list of major South African online retailers.
Identifies the absolute best price and the corresponding link.
Sends (tries to send) a single, consolidated daily email summary of all significant updates (price drops, stock changes, new best prices found) to the user, preventing notification overload.
Logs its actions and any errors directly back into the Google Sheet for transparency and future review.
This hybrid approach leverages the agent's automation power for repetitive tasks while maintaining human control over critical initial data and error resolution, providing both efficiency and accuracy.
Demo
How I Used Runner H
Natural Language Automation: The entire workflow is defined through a single, comprehensive natural language prompt.
(Surfer H) & Autonomous Search:Autonomously search for products: When a Product Name is added to the sheet without a Product Found URL, Surfer H performs a Google Search across specified retailers to find and suggest the most relevant product page. This is the first "human-in-the-loop" point, where the user reviews the agent's suggestion in the Google Sheet.
Navigate to any e-commerce website.
Identify and extract dynamic data (prices, stock status) from various website layouts.
Visit multiple links from search results to compare prices across different retailers.
Multi-Step Task Completion: Runner H excels at orchestrating a series of complex, interdependent steps:
Reading data from an external source (Google Sheet).
Performing conditional actions (initial product search if URL is empty).
Executing web browsing and data extraction tasks.
Performing data comparison and logical decision-making (price drop >10%, stock change, best price found).
Updating an external data source (Google Sheet) with new information. This is where the agent logs its progress, including suggested URLs and any errors, for human review.
(Tries to) Composing and sending a consolidated email notification.
Managing its own state (via Notification Sent (Last Event) column) to prevent redundant notifications.
Instructions for Replication:
To replicate this semi-autonomous wishlist monitoring workflow, follow these steps:
Prepare Your Google Sheet:
Click this link to make a copy of the pre-formatted Google Sheet template:
🔗 Click here to access your Wishlist Monitoring Google Sheet TemplateCrucially, set the sharing settings for your copied Google Sheet to "Anyone with the link can edit". This allows Runner H to update the sheet.
In your copied sheet, fill in the Product Name for each item you want to monitor.
Optionally, fill in Desired Price (R).
Fill in the Notification Email for each product. If left blank, notifications will not work (struggles to work even if available).
Leave the Product Found URL column blank for any new products you add by name. The agent will populate this.
Leave all other (Agent Populated) columns blank; Runner H will populate them.
Copy the full URL of your Google Sheet from your browser's address bar.
Access Runner H (H-Studio):
Log in to your Runner H account.
Give access to Google Sheets & Gmail in Connections.
Submit the Workflow Prompt:
Provide access to the sheets tool and Gmail.
Submit the prompt below into the input area in Runner H's H-Studio.
Paste your Google Sheets link in step 1 of the Workflow steps.
Start of the prompt
Objective: Operate as a fully autonomous agent to continuously monitor a list of products (identified by name) from a Google Sheet for price changes and stock availability, find the best price across multiple retailers, and send a single, consolidated daily email summary of all relevant updates.
Workflow Steps:
Access Google Sheet: Connect to the Google Sheet located at [Google Sheets Link]. Read all rows from this sheet.
Initialize Daily Update List: Create an empty list to store details of all products that have a significant update during this run.
Iterate Through Products: For each row (product) in the Google Sheet: a. Initial Product Identification (if 'Product Found URL' is empty): * IF the
Product Found URL (Agent Populated)
column for the current row is empty: * Perform a Google Search using theProduct Name
from the current row, prioritizing results from major South African retailers (Takealot, Makro, Dion Wired, BobShop, Superbalist, Amazon.sa, Incredible Connection, Game, Loot.co.za, Hirsch's, Builders Warehouse, Mr Price, Sportscene, Everyshop, House & Home). * Automatically select the first search result URL that appears to be a direct product page from one of the prioritized retailers. Prioritize results from the most reputable and common e-commerce sites. * Update theProduct Found URL (Agent Populated)
column in the Google Sheet for this row with the selected URL, for each product. * IF no suitable product page is found after a thorough search (e.g., no relevant results from prioritized retailers), update theNotification Sent (Last Event)
column for this row with "Error: Product not found during initial search on [Current Date/Time SAST]". Log this error internally and proceed to the next product without attempting further monitoring for this specific item in this run. for each product do the following: b. Navigate & Extract Product Data (using 'Product Found URL'): * Using Surfer H, navigate to theProduct Found URL (Agent Populated)
specified in the current row. * Implement robust extraction: Extract the current price of the product and its stock status (e.g., "In Stock", "Out of Stock", "Low Stock"). Handle common variations in how price and stock are displayed on different websites. * Update theCurrent Price (R)
,Last Checked
(with current date and time in SAST), andStock Status
columns in the Google Sheet for this row. * IF there's an error during this step (e.g., page not found, price element missing), updateNotification Sent (Last Event)
with "Error: Failed to extract data from primary URL on [Current Date/Time SAST]". Log this error internally and proceed to the next product. c. Check for Cheaper Alternatives: * Perform a Google Search for theProduct Name
along with the names of other major South African retailers (Takealot, Makro, dion wired, BobShop, Superbalist, Amazon.sa, Incredible Connection, Game, Loot.co.za, Hirsch's, Builders Warehouse, Mr Price, Sportscene, Everyshop, House & Home). * Analyze the search results and visit promising links to find the same product. * Identify the lowest price for the product found on any of these alternative websites. * IF a price lower than theCurrent Price (R)
(from the primary URL) is found: * UpdateCheaper At (Website)
,Cheaper At (Price) (R)
, andCheaper At (Link)
columns in the Google Sheet for this row with the details of the best alternative found. * ELSE (if no cheaper alternative or if the primary URL is still the best): * Clear these three columns (Cheaper At (Website)
,Cheaper At (Price) (R)
,Cheaper At (Link)
). d. Determine Notification Need & Add to Daily Update List: * Compare the newly extractedCurrent Price (R)
with theDesired Price (R)
(if provided) and the price from the previous run (which would be theCurrent Price (R)
from the last time the agent ran before this update). * Calculate the percentage price drop from the previous run's price. * Check if theStock Status
has changed from "Out of Stock" to "In Stock". * Check if a newCheaper At (Website)
andCheaper At (Price) (R)
were identified, or if the best alternative price is significantly lower (e.g., by more than 5%) than the current primary URL price. * Construct aCurrent Event Description
string (e.g., "Price dropped to R18999", "Back in stock", "Cheaper alternative found at Makro for R18500"). * Retrieve theNotification Sent (Last Event)
from the current row. * IF (a significant price drop (>10% from previous run's price) OR a stock change (Out of Stock to In Stock) OR a new, significantly cheaper alternative is found) AND theCurrent Event Description
is different from theNotification Sent (Last Event)
: * Add theProduct Name
,Current Status on Primary Site
(Price, Stock, Link),Details of Change
, andBest Price Found (if applicable)
to theDaily Update List
. * Update theNotification Sent (Last Event)
column in the Google Sheet for this row with theCurrent Event Description
and the current date/time.Send Consolidated Daily Email Summary: **Email Subject:* "Daily Wishlist Summary - [Current Date SAST]" * Email Body: "Hello, Here is your consolidated daily wishlist update for [Current Date SAST]: [For each item in the Daily Update List, format as follows]: --- Product: [Product Name] Current Status on Primary Site: Price: R[Current Price (R)] Stock: [Stock Status] Link: [Primary Found URL (Agent Populated)] Details of Change: [Describe the specific change, e.g., "Price dropped by X% from its previous check.", "Item is now back in stock!", "Found a new best price."] Best Price Found (if applicable): [IF 'Cheaper At (Website)' was populated for this item]: Website: [Cheaper At (Website)] Price: R[Cheaper At (Price) (R)] Direct Link: [Cheaper At (Link)] --- [If there were any errors during the run, add a section like this]: Errors/Issues Encountered: Please check the 'Notification Sent (Last Event)' column in your Google Sheet for details on any products that could not be processed fully. Happy shopping! Your Autonomous Wishlist Agent"
Schedule: Run this entire workflow once every day at 06:00 AM SAST (South African Standard Time).`
End of the prompt
Use Case & Impact
Wishlist monitoring, executed by such an agent, offers significant real-world benefits. Here's a breakdown:
🛒 Smart Shopping for Big Purchases: For high-value items like electronics, appliances, furniture, or even cars, price fluctuations can be significant. The agent ensures you never miss a substantial discount.
🎁 Effortless Gift Planning: No more last-minute scrambles! The agent can monitor gift ideas for birthdays, anniversaries, or holidays throughout the year, ensuring you buy thoughtful presents at the best possible price.
✈️ Travel Gear & Experiences: From luggage and outdoor equipment to even flight deals (if the agent can extend its capabilities to travel sites), you can get alerts when prices drop.
📚 Hobby & Niche Item Acquisition: For specialized equipment (e.g., photography gear, musical instruments, collector's items), the agent can track prices in niche online stores or marketplaces, which might not be covered by standard price comparison sites.
Virtually anyone who shops online can benefit, but certain groups would find it particularly transformative:
💰 The Budget-Conscious Shopper: Someone who wants to maximize savings on every purchase, no matter how small. They'll appreciate knowing exactly when to strike for the best deal.
🕰️ The Time-Poor Professional/Parent: Individuals with demanding schedules who simply don't have the time to scour multiple websites daily for deals. The agent automates this tedious process.
The Runner H AI Agent upgrades how people currently approach shopping and saving.
- No more repeatedly visiting sites to see if an item's price has changed. This is the biggest time-saver.
- You get factual, real-time data on price history and current market conditions, leading to smarter buys.
- With Runner H is constantly monitoring your desired items, you'll never miss a limited-time sale or a stock replenishment.
- Runner H can search and compare prices across a wider range of retailers than any human could manually, booting your odds of finding the best deal.
Social Love
I post on X. https://x.com/directors_comp
Top comments (0)