Instructions, tools, skills, files, and models

Define what an agent may do before giving it access to a system.

• Agent Instructions: the goal, boundaries, and expected output.

• Agent Tools: the actions it can call; review credentials and write permissions.

• Agent Skills: reusable capabilities for a task.

• Agent Files: context the agent may read.

• AI Models: choose a model available to your plan and deployment.

Start with a narrow task and the minimum necessary tools. Test the agent with records it should and should not access before connecting it to users or workflows.