Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.eluu.ai/llms.txt

Use this file to discover all available pages before exploring further.

Some work has too many strands for one chat thread. Researching ten companies, exploring three different solutions, planning a project across four functions. When that happens, your colleague can dispatch sub-agents — focused workers that go off, do one piece, and come back with a clean answer. You stay in the main chat. The sub-agent does the heavy lifting in the background.

​
Why use them

The main reason: don’t bloat the conversation. If you ask a colleague to research ten companies in one go, all the half-formed thoughts and dead-end searches stay in the chat history. Memory gets cluttered. Future answers get fuzzier. Send those ten research jobs to ten sub-agents instead. Each one does its piece in isolation, comes back with a tight summary, and the main chat stays clean.

​
When to ask for them

A few patterns where sub-agents help:
  • Anything in parallel. Researching multiple companies, drafting outreach for ten leads, summarising five long documents. Each one a sub-agent.
  • Exploring options. Considering three different angles on a strategy — give each angle a sub-agent and compare the outputs.
  • Verification. Have a sub-agent independently verify the work the main colleague did. They don’t see each other’s reasoning, so the verifier catches mistakes the main one missed. (More on this in verification techniques.)
  • Long-running research. A sub-agent that goes deep, while you keep working with the main colleague.

​
How to ask

Just tell the colleague.
Use sub-agents to research these five companies in parallel. One per company. Have each one come back with the same shape: company, recent news, two potential pain points, named contact.
For this plan, use a sub-agent to verify the numbers in section 3 independently.
Send a sub-agent off to read all our recent customer call notes and summarise patterns. Keep working on the report meanwhile.
The colleague knows which kind of sub-agent to use — research, planning, exploration, or general purpose — and dispatches it.

​
What you’ll see

The sub-agent shows up as a card in the chat: what it’s doing, when it finishes. You can expand to see the work, but you don’t have to. The main colleague pulls in just the result and keeps going.

​
Where to next

Sub-agents for verification

The technique that makes scheduled jobs trustworthy.

Skills

Repeatable multi-step workflows.