Instructions agents are internal agents powered by a description you provide of their personality and what they do. They live entirely inside Dart and spring into action the moment work is assigned to them.
Creating an Instructions agent
Go to Agents in the left navigation bar and select Create a custom agent
Give the agent a Name and Icon
Choose Instructions as the type
Select the AI model that powers it (Auto, Claude, ChatGPT, or Gemini)
Write the agent's guidelines and configure any automated status changes
Writing effective guidelines
Your agent's guidelines are its rulebook. Be highly specific about the steps it should take when it receives an assignment. Example use cases:
Bug & issue triage: Scan new bug reports for reproduction steps and logs, check for duplicates, set priority, and summarize for the team
Integration & payload parsing: Rurn raw webhook JSON into a clean Markdown table and route it to the right queue
QA & review: Verify subtasks and deliverables, mark Done or set Blocked with a comment
Standup prep: Summarize yesterday's updates, flag stuck tasks, and create a summary doc
Onboarding: Generate onboarding tasks and a welcome doc when someone joins
Sprint cleanup: Roll over incomplete tasks and summarize what shipped