What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY  
────────────────────────────────────────────────────────
“MerchDataCollector” performs live searches for best-selling T-shirt listings on Merch by Amazon (or Amazon.com), collects those listings, and filters out any that have fewer than 3 reviews or that violate content policies.

B) FINAL TASK OUTPUT  
────────────────────────────────────────────────────────
A structured list of filtered t-shirt listings (in memory), each containing product details such as title, bullet points, price, reviews, rating, etc.  
Label: [filtered-product-listings]

C) SUBAGENT INPUT  
────────────────────────────────────────────────────────
• Category name (e.g., “Sports,” “Funny,” “Holiday,” etc.)  
• Time frame or general instructions on “last 30 days” trending data

E) SUBAGENT TASK SUMMARY (Step-by-Step Flow)  
────────────────────────────────────────────────────────
1) Receive category and timeframe instructions from the higher-level agent.  
   └─ Input: {category: “Sports”, timeframe: “Last 30 days”}

2) Skill #224 - Oracle Ask A Question  
   └─ Purpose: Find relevant URLs on Amazon or Merch by Amazon showing best-selling T-shirts in the specified category.  
   └─ Input: “Please provide up to 3 relevant Amazon URLs for top trending T-shirts in [CATEGORY] over the last 30 days.”  
   └─ Output: Live or up-to-date web result containing 1–3 suggested URLs.

3) Parse the Oracle’s output and identify valid Amazon URLs.  
   └─ (Internal step: read the returned text from #224, pick relevant link(s).)

4) Skill #225 - Crawl 1x Amazon URL (Iterate if multiple URLs needed)  
   └─ Purpose: For each identified URL, crawl the Amazon page(s) to extract structured product data.  
   └─ Input: “Crawl this Amazon URL: <URL>. Extract all relevant T-shirt listing data such as title, bullet points, price, reviews count, rating, BSR.”  
   └─ Output: Structured data about T-shirt listings on that page.

5) Combine all extracted listing data (if multiple URLs).  
   └─ (Internal step: merge into a single dataset/collection of raw T-shirt listings.)

6) Skill #223 - Powerful LLM Prompt-to-Text Response  
   └─ Purpose: Filter out any listings that have fewer than 3 reviews or contain content that violates Amazon policies.  
   └─ Input: (a) The T-shirt listing dataset from step 5, (b) Instructions to remove listings with fewer than 3 reviews or flagged content.  
   └─ Output: The final collection of T-shirt listings meeting the subagent’s criteria.

7) Provide final subagent output.  
   └─ Label: [filtered-product-listings]

F) SILOS  
────────────────────────────────────────────────────────
• SILO 1: “Find Relevant Links”  
  - Skill: #224 Oracle Ask A Question  
  - Goal: Obtain up to 3 relevant Amazon URLs showing best-selling T-shirt listings in the specified category.

• SILO 2: “Extract Product Data”  
  - Skill: #225 Crawl 1x Amazon URL (repeated if multiple URLs)  
  - Goal: Get structured data (title, bullet points, price, reviews, rating, BSR).

• SILO 3: “Filter Listings”  
  - Skill: #223 Powerful LLM Prompt-to-Text Response  
  - Goal: Remove listings with <3 reviews or TOS violations. Return the final [filtered-product-listings].

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
────────────────────────────────────────────────────────
"TShirtRanker" takes a list of valid t-shirt listings (already filtered by another subagent) and determines the top 5 products based on BSR, review count, average rating, and price.

B) FINAL TASK OUTPUT  
────────────────────────────────────────────────────────
[top5-ranked-listings] → A structured list (array) of exactly five (5) t-shirt product listings, each assigned a rank (1–5).

C) SUBAGENT INPUT  
────────────────────────────────────────────────────────
• A list/array of filtered t-shirt listings containing (at minimum) the following fields:  
  - title  
  - bsr  
  - price  
  - reviews.count  
  - reviews.rating

E) SUBAGENT TASK SUMMARY (Step-by-Step Flow)  
────────────────────────────────────────────────────────
1) SUBAGENT INPUT  
   Description: The subagent receives the filtered listings array from the previous subagent (“MerchDataCollector”).

2) SKILL #223 - Powerful LLM Prompt-to-Text Response  
   Description: Rank and select the top 5 listings.  
   Input:  
      • The filtered listings array (JSON or similarly structured text).  
      • Clear instructions on how to sort, for example:  
        "Sort all items primarily by Best Sellers Rank (BSR) ascending (lowest BSR first). If any ties occur, prioritize higher average rating. If still tied, higher review count. If still tied, lower price. Then select the top 5. Return them as an array with the fields and rank appended."  
   Output:  
      • A textual or JSON-like structure with the top 5 t-shirt listings, each labeled “rank”: 1–5.

3) SUBAGENT FINAL OUTPUT  
   Description: Return the final set of top 5 ranked t-shirt listings with rank appended.  
   Output:  
      • [top5-ranked-listings]

F) SILOS  
────────────────────────────────────────────────────────
• SILO 1: "Load & Sort"  
  - The subagent receives the list of filtered listings as input and calls SKILL #223 to determine sorting order.

• SILO 2: "Prepare Top 5"  
  - The subagent extracts the top 5 from the sorted list, appends rank (1–5), and outputs [top5-ranked-listings].

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
"JsonFormatter" takes the finalized top 5 t-shirt listings (along with their category and the date analyzed) and transforms them into a valid JSON string conforming exactly to the required schema.

B) FINAL TASK OUTPUT  
The subagent’s final output is a single JSON string in memory that follows the schema:  
{
  "category": "",
  "date_analyzed": "",
  "products": [
    {
      "rank": 1,
      "title": "",
      "bullet_points": [
        "",
        ""
      ],
      "description": "",
      "price": "",
      "reviews": {
        "count": "",
        "rating": ""
      },
      "primary_keywords": []
    },
    ...
  ]
}

C) SUBAGENT INPUT  
• The set of five top ranked t-shirt listings (each containing title, bullet points, description, price, reviews count, rating, etc.).  
• The category name (string).  
• The date analyzed (string).  
• Instructions specifying the exact JSON schema and required data fields.

E) SUBAGENT TASK SUMMARY  
1) Receive the subagent input (top 5 listings, category, date analyzed, JSON schema requirements).  
2) Use skill #223 (Powerful LLM Prompt-to-Text Response) to format the data into the specified JSON structure.  
   - Input to #223:  
       - A prompt containing the aggregated listing data, the category, the date_analyzed, and the exact schema instructions.  
   - Output from #223:  
       - A single well-formed JSON string that includes all required fields.  
3) Return that JSON string as the subagent’s final output in memory, ready for use by subsequent subagents (e.g., the “FinalOutputWriter”).

F) SILOS  
• Silo 1 – Prepare Data:  
  - No explicit skill used here; just gather the top 5 listings, category, date_analyzed, and schema instructions from prior subagents or inputs.  
• Silo 2 – JSON Construction:  
  - Skill #223 (Powerful LLM Prompt-to-Text Response)  
    → Input: “Construct JSON using this data and these instructions.”  
    → Output: The fully formatted JSON string.  
• Silo 3 – Output Delivery:  
  - The subagent returns the finalized JSON string in memory as [final-json-in-memory].

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY  
"FinalOutputWriter" is responsible for taking the finalized JSON string (with the top 5 t-shirt listings) and delivering a single JSON file named “top5_tshirts.json.”

B) FINAL TASK OUTPUT  
A JSON file named “top5_tshirts.json” whose content is the completed JSON data (formatted according to the specified schema) containing the top 5 t-shirt listings.

C) SUBAGENT INPUT  
• The final JSON string produced by the previous subagent (“JsonFormatter”), which already includes all fields in the required schema (category, date_analyzed, products, etc.).  
• Instructions to produce the output file named “top5_tshirts.json.”

E) SUBAGENT TASK SUMMARY  
Below is the step-by-step flow of how “FinalOutputWriter” will create the final JSON file:

1. Receive the completed JSON string from Subagent 3’s output.  
2. Invoke Skill #185 (Write Text (Or Copy) From Inputted Text) with the following parameters:  
   • INPUT:  
     – The JSON string and a simple instruction to produce “top5_tshirts.json.”  
   • ACTION:  
     – Write or copy the JSON exactly into a file named “top5_tshirts.json.”  
   • OUTPUT:  
     – The generated text (which is effectively the entire JSON content).  
3. The final subagent output will be named [top5_tshirts-json-file], representing the “top5_tshirts.json” file containing the top 5 t-shirt data.

F) SILOS  
Because this subagent’s sole function is to produce the single JSON file, it effectively has one silo:  
• Silo 1: Convert Final JSON String into “top5_tshirts.json”  
   – Input: Final JSON string.  
   – Process: Skill #185 (Write Text / Copy JSON).  
   – Output: “top5_tshirts.json” file.

4 Template & Links

Expand Flow

A) SUBAGENT SUMMARY  
"Subagent #5" is responsible for identifying recurring keywords and themes among the already-selected top 5 products (which have passed all filters and ranking). It reviews each product’s title, bullet points, and description, then outputs a concise set of recurring keywords or top themes.

B) FINAL TASK OUTPUT  
A text-based list (or short summary) of the recurring keywords or themes found across the final 5 product listings. The subagent’s output can then be appended into the final JSON if desired.

C) SUBAGENT INPUT  
• The final curated list of 5 products (with title, bullet points, descriptions, etc.) from the earlier subagents.  
• Optional: any previous notes on themes or partial keyword lists (if available).

E) SUBAGENT TASK SUMMARY  
Below is the step-by-step breakdown of how Subagent #5 accomplishes its task:

1. Receive Final Top 5 Listings  
   - Input: A structured list of 5 products containing at least: title, bullet points, and product description.

2. Identify Keywords & Themes (Skill #223: Powerful LLM Prompt-to-Text Response)  
   - The subagent uses an LLM to parse the textual fields (titles, bullet points, descriptions).  
   - Prompt instructs the LLM to:  
     • Extract repeated words or phrases.  
     • Exclude generic or irrelevant words (like "shirt," "Amazon," "design," “funny”—unless they truly stand out as themes).  
     • Return a concise ordered list (or short commentary).  
   - Output: A short textual summary containing the top recurring keywords/themes.

3. Final Output Collation  
   - The subagent collates the recurring keywords or theme summary into one final text snippet (e.g., “Recurring_Keywords_And_Themes”), ready to be added to the final pipeline or stored separately.

F) SILOS (If Needed)  
• Silo 1: “LLM Analysis”  
  - Exclusive use of Skill #223 to parse through the final 5 listings and return a short summary of recurring keywords or themes.

No additional silos required; the subagent is focused on a single main function: analyzing text to find recurring keywords/themes.

5 Template & Links

Expand Flow

A) SUBAGENT SUMMARY  
"RecurringKeywordAnalyzer" will take in a set of valid t-shirt listings (with titles, bullet points, descriptions, etc.) and identify any recurring keywords or themes appearing across these listings, returning a concise summary of these keywords along with their relative frequency or importance.

B) FINAL TASK OUTPUT  
A text-based report (string) listing the recurring themes or keywords found in the provided t-shirt listing data, optionally including basic frequency counts or short commentary on each keyword.

C) SUBAGENT INPUT  
• A list of t-shirt entries in a structured data format (e.g., JSON array) containing:  
  – Title  
  – Bullet points  
  – Description  
  – Price  
  – Reviews (count + rating)  
• A brief explanation or requirement regarding the level of detail needed for the keyword/theme analysis (optional).

E) SUBAGENT TASK SUMMARY  
1) The subagent starts by accepting the structured list of t-shirt data from a previous subagent.  
2) It then prompts a powerful LLM (Skill #223 - Powerful LLM Prompt-to-Text Response) to:  
   – Read through each listing’s title, bullet points, and description.  
   – Extract words or short phrases appearing with notable frequency or conceptual overlap.  
   – Group, rank, or summarize them by significance.  
   – Produce a final text describing the recurring keywords and themes found.  
3) The subagent outputs the final text-based list/report with any relevant frequency data, concluding the subagent’s work.

Visually, the flow can be represented as:

[Subagent #6 input: structured listings]  
          ▼  
    (Skill #223) → “Analyze listing text to detect recurring keywords or themes”  
          ▼  
   [Text-based output containing discovered keywords/themes]

F) SILOS  
• PREPARE INPUT SILO  
  – Confirm the data structure (list of t-shirt entries with relevant fields) is valid.

• KEYWORD EXTRACTION & ANALYSIS SILO  
  – Use Skill #223 (Powerful LLM Prompt-to-Text Response) to parse and analyze t-shirt listing text content.  
  – Extract recurring keywords or themes (e.g., sports, funny, holiday references, etc.).

• OUTPUT SILO  
  – Return a single text report enumerating the primary recurring keywords/themes along with basic frequency or context if desired.