What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY  
"AmazonScraper" crawls a single Amazon Best Sellers page, gathers ~5 products, and selects the highest-priced product among them.

B) FINAL TASK OUTPUT  
A JSON/dict object containing:  
{  
  "product_name": "...",  
  "product_price": "...",  
  "product_link": "...",  
  "product_details": "..."  
}

C) SUBAGENT INPUT  
{amazon-category-url} (the URL of the Amazon best sellers page for the chosen category).

E) SUBAGENT TASK SUMMARY  
1) Input → Skill #225 - Crawl 1x Amazon URL  
   • INPUT: {amazon-category-url} plus the instruction: "Extract ~5 best sellers; provide product name, price, link, and brief details."  
   • OUTPUT: A summarized list (in JSON or similar text format) of the ~5 best sellers.

2) Skill #223 - Powerful LLM Prompt-to-Text Response  
   • INPUT: The summarized list from step 1 and a prompt requesting selection of the highest-priced product.  
   • OUTPUT: JSON/dict with "product_name", "product_price", "product_link", "product_details".

3) SUBAGENT FINAL OUTPUT: [highest-priced-product-data]

F) SILOS  
No additional silos are required. The two skills form a straightforward linear sequence, culminating in the highest-priced product data.

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
"ReviewSummarizer" takes the name (and basic info) of the highest-priced best-selling product and produces a concise written review by researching Google/YouTube sources and then condensing that research into a short, user-friendly write-up.

B) FINAL TASK OUTPUT  
[short-product-review] → A concise text review (a few paragraphs) describing the product’s main features, pros, and cons.

C) SUBAGENT INPUT  
[highest-priced-product-data]  
This is a JSON/dict containing (at minimum) the following keys:  
• "product_name"  
• "product_price"  
• "product_link"  
• "product_details"

D) SUBAGENT FINAL OUTPUT  
[short-product-review]  
A short text review (string).

E) SUBAGENT TASK SUMMARY  
Below is the step-by-step chain of skills that Subagent 2 performs to achieve its final output:

1) #216 - Research Topic (Or Keyword) Deeply  
   — INPUT: The "product_name" from [highest-priced-product-data], with instructions to search for product reviews and feedback on Google and YouTube.  
   — OUTPUT: A ~1–3 paragraph researched summary of pros, cons, notable features, etc.

2) #185 - Write Text (Or Copy) From Inputted Text  
   — INPUT: The text output from Step 1, with instructions to produce a concise, user-friendly product review (simply recheck and condense as needed).  
   — OUTPUT: The final short review text, [short-product-review].

F) SILOS  
• SILO 1: Gather Research  
   - Use Skill #216 (Research Topic) with the product name.  
• SILO 2: Generate Final Review  
   - Use Skill #185 (Write Text) to condense the research into a concise review.

This completes the workflow for Subagent 2: "ReviewSummarizer."

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent (“ImageFinder”) locates and returns up to three relevant product image URLs based on the product name from the highest-priced Amazon product data.

B) FINAL TASK OUTPUT  
[product-image-urls] (a list/array of up to three URLs of the most relevant product images)

C) SUBAGENT INPUT  
[highest-priced-product-data] – the JSON/dict containing keys such as “product_name”, “product_price”, “product_link”, and “product_details”, all describing the chosen product.

E) SUBAGENT TASK SUMMARY  
1) From [highest-priced-product-data], identify the “product_name”.  
2) Use #187 (Find Relevant Images with Brave Search) to gather 5 relevant image URLs for that product name.  
3) Use #223 (Powerful LLM Prompt-to-Text Response) to filter those 5 URLs down to the top 3 URLs (or fewer, if fewer than 3 are genuinely relevant).  
4) Return those up to three final image URLs as [product-image-urls].

F) SILOS

• SILO 1: IMAGE SEARCH  
   – Input: (product_name from [highest-priced-product-data])  
   – Action: #187 - Find Relevant Images with Brave Search  
   – Output: List of 5 image URLs

• SILO 2: FILTER IMAGES  
   – Input: The 5 URLs from SILO 1  
   – Action: #223 - Powerful LLM Prompt-to-Text Response (instruct the model to select only the top 3 most relevant image URLs)  
   – Output: Final [product-image-urls] (up to three URLs)

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY  
"HTMLComposer" takes the product info, review text, and image URLs, then outputs a single valid HTML file (as a string) ready for publication.

B) FINAL TASK OUTPUT  
A single HTML file/string that includes:  
• The product name, link, and key details (from AmazonScraper)  
• The short review text (from ReviewSummarizer)  
• 1–3 embedded product images (from ImageFinder)

C) SUBAGENT INPUT  
• [highest-priced-product-data] (JSON or dict with product_name, product_price, product_link, product_details)  
• [short-product-review] (string of review content)  
• [product-image-urls] (array/list of 1–3 image URLs)

E) SUBAGENT TASK SUMMARY (Step-by-step Action Flow)  
1) #185 - Write Text (Or Copy) From Inputted Text  
   DESCRIPTION: Combine the provided product data, review text, and image URLs into a properly structured HTML document.  
   INPUT: A short prompt/instruction that says, for example:  
   “Use these inputs to create a final HTML review article:  
     • Product Data: {{product_name, product_price, product_link, product_details}}  
     • Review: {{short-product-review}}  
     • Image URLs: {{product-image-urls}}  
   Make sure to embed the images and include the product link as a hyperlink. Return the HTML as a string.”  
   OUTPUT: A single HTML string that meets the final output requirements.

F) SILOS  
For "HTMLComposer," there is only one skill/action in its silo. The subagent gathers all inputs and invokes skill #185 once to generate the final HTML.