What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
"StepExtractorAgent" reads the time-stamped transcript, determines exactly six key steps with one representative timestamp each, and outputs a structured list containing those six steps.

B) FINAL TASK OUTPUT  
A single structured data object, [six-steps-list], containing exactly six steps. Each step includes:  
• A short title or label  
• A brief description of what happens in that step  
• A single timestamp (HH:MM:SS or similar)

C) SUBAGENT INPUT  
• [time-stamped-transcript]: The complete transcript of the video (with timing markers).

E) SUBAGENT TASK SUMMARY  
Below is the chain of actions for Subagent 2 to produce [six-steps-list]:

1) Receive [time-stamped-transcript].  
2) Invoke Skill #223 (Powerful LLM Prompt-to-Text Response) with a prompt instructing the LLM to:  
   • Read the transcript thoroughly.  
   • Identify/summarize exactly six major steps or segments from the transcript (regardless of how many steps the narrator actually mentions).  
   • For each identified step, pick one relevant timestamp that best corresponds to the start (or key demonstration) of that step.  
   • Return a structured list of these six steps, with each step containing a short title, a brief descriptive text, and a single concise timestamp.  
3) The final output from this single skill call is [six-steps-list].

Therefore, the subagent’s entire process only needs one skill call to #223, ensuring we get the final structured list in the desired format.

F) SILOS  
This subagent is straightforward and does not require distinct silos. All work is done in a single skill call #223, which produces the final consolidated [six-steps-list].

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
"This subagent, the ‘FrameGrabberAgent,’ extracts exactly six image frames from the input video at the six timestamps determined by the previous Subagent, then collates those six image URLs into a single structured output."

────────────────────────────────────────────────

B) FINAL TASK OUTPUT  
"[six-image-urls]" → A single structured text (or JSON-like) object containing exactly six image references (URLs) corresponding to the six timestamps.

────────────────────────────────────────────────

C) SUBAGENT INPUT  
1. {variable1} (the MP4 video URL)  
2. [six-steps-list] (the structured list of six steps with timestamps)

────────────────────────────────────────────────

E) SUBAGENT TASK SUMMARY

Below is the exact chain of skills/actions for Subagent 3:

1) Receive the subagent’s two required inputs:  
   - MP4 video URL: {variable1}  
   - Six-step list: [six-steps-list], which includes each step’s timestamp.

2) For each of the six steps, retrieve the timestamp from [six-steps-list], then call Skill #202 - Extract Thumbnail Images Of MP4 Video, passing in:  
   - The MP4 video URL ({variable1})  
   - A short prompt or instruction requesting extraction of a video frame near the step’s timestamp.  
   Note: Skill #202 can return up to 10 thumbnail images. We only keep/identify the single best or first returned image that corresponds to that timestamp for each step.

3) Collect the resulting six images (one per step). We will have six unique image URLs after all calls are complete.

4) Finally, call Skill #223 - Powerful LLM Prompt-to-Text Response to assemble these six image URLs into one final structured output. (For example, a JSON array or list mapping step index → image URL.)

5) The subagent’s final output ([six-image-urls]) is the single text-based structure containing the six image references, which will be used by the next subagent.

────────────────────────────────────────────────

F) SILOS

SILO 1: Extract Frames For Each Step  
• Action Repeated 6 Times:  
  → Skill #202 (Extract Thumbnail Images Of MP4 Video)  
  Input: {variable1} and a brief prompt referencing the desired timestamp  
  Output: Usually up to 10 images. We pick/label the single best image.

SILO 2: Collate Results  
• Action Once:  
  → Skill #223 (Powerful LLM Prompt-to-Text Response)  
  Input: The six chosen image URLs  
  Output: [six-image-urls] (a unified textual/JSON structure containing the 6 image URLs)

────────────────────────────────────────────────

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY  
This subagent (“HTMLBuilderAgent”) uses the final six-step list, six images, and the original transcript to produce a single polished how-to HTML article with exactly six steps—each step containing its own image—ready for immediate publication.

B) FINAL TASK OUTPUT  
[final-html-article] → A finalized HTML file (text format) containing a title, a brief introduction, six steps (each with a heading, short description, and an embedded image), and an optional conclusion.

C) SUBAGENT INPUT  
• [time-stamped-transcript] – The raw transcript text with timestamps.  
• [six-steps-list] – A structured list containing six steps with their titles, short descriptions, and timestamps.  
• [six-image-urls] – A structured list containing six image URLs (or filenames), each corresponding to one step.

D) SUBAGENT TASK SUMMARY  
1. Receive the three inputs: [time-stamped-transcript], [six-steps-list], [six-image-urls].  
2. Use Skill #190 - “Write or rewrite text based on instructions” to take these inputs and assemble them into a single coherent HTML file. This should include:  
   • A short overall introduction (1–2 paragraphs) derived from or inspired by the transcript/theme.  
   • Exactly six steps, each with:  
       – A descriptive heading.  
       – A concise explanation/description (rewritten or summarized from [time-stamped-transcript] if needed).  
       – An embedded image tag <img src="IMAGE_URL" ...> referencing the matching URL from [six-image-urls].  
   • An optional closing or summary paragraph.  
3. The output of Skill #190 is the final HTML article string.  
4. Subagent returns [final-html-article] containing the finished HTML.

E) SILOS  
No separate silos are required; this subagent is a straightforward single-skill process:

[time-stamped-transcript], [six-steps-list], [six-image-urls]  
   ↓  
( Skill #190: Write or rewrite text based on instructions → Outputs final HTML )  
   ↓  
[final-html-article]