Home › Tutorials › Autonomous Agents
Autonomous Agents

Talking To & Controlling Agents

#077 0:20 Part of the Rysh video series

@name to prompt it, @@name to command it.

Talking To & Controlling Agents — 0:20 walkthrough

What you'll see

  1. Launch rysh; double-Escape into rysh mode; spawn code-reviewer.
  2. @code-reviewer what are the top three things you check in a pull request? -- single @ sends a prompt.
  3. @@code-reviewer deactivate -- double @@ sends a control command; deactivate keeps state but ignores prompts.
  4. @@code-reviewer activate -- bring it back.
  5. ##agent delete code-reviewer -- remove it entirely.

Commands shown

rysh
##agent spawn code-reviewer 'You review code carefully and concisely.'
@code-reviewer what are the top three things you check in a pull request?
@@code-reviewer deactivate
@@code-reviewer activate
##agent delete code-reviewer

Keys used

EnterEscape

Transcript

0:00At-name to prompt it, double-at-name to command it.

0:03Launch rysh, drop into rysh mode, and spawn a reviewer to work with.

0:10A single at-sign plus the name sends a prompt. The agent runs it with its full toolbelt and reports back.

0:22Two at-signs send a control command instead of a prompt. Deactivate pauses the agent -- it keeps its state but ignores prompts.

0:31Activate brings it right back to life.

0:38And ##agent delete removes it entirely when you're done. Prompt, pause, resume, delete -- you're always in control.

Key takeaway
@name <prompt> prompts an agent; @@name <stop|activate|deactivate> controls it; ##agent delete removes it.