Collaborative bot building guide (bot chaining)

Goal

Build a chain of custom bots that work together on a multi-step task — each bot handles one step and passes its output to the next. (BoodleBox calls these "skeletons of agents.")

The idea

Instead of one bot doing everything, break a problem into steps and build a dedicated bot for each step. In a single chat, each bot can see what the previous bots produced, so the work builds up naturally.

How to build a chain

1. Decompose the problem. Use systems-thinking questions — inputs, outputs, boundaries, constraints, feedback loops — to decide how many bots you need and what each one does. (Tip: ask a strong model to map this out for you: "Break this down into bots using systems thinking — what does each bot do?")

2. Build each bot in the Bot Builder with explicit instructions (see Designing effective bot instructions):

   • Input: "You'll receive [structured data] from [previous bot]."
   • Role: "Your job is [X]."
   • Output: "Produce [structured format] for [next bot]."

3. Run the chain in one chat. Upload the problem, then @-mention Bot 1 and say "do it"; @-mention Bot 2, which reads Bot 1's output; continue through the chain. Because every bot sees the conversation above it (bot stacking), little or no extra prompting is needed.

Great for the classroom

Have students design their own bot chains for a case study — it teaches decomposition and systems thinking.

Questions?

Reach out through our contact form and we'll get back to you. (See also How to build a custom bot and How to use more than one AI in the same chat.)