What is the role of tool permissions for an agent?
Why C is the answer
They limit what the agent is able to do in its environment
Selecting and scoping tools is the first line of control. An agent cannot misuse a tool it does not have.
In detail
- Tool permissions limit what an agent can do in its environment.
- Grant tools by job.
Use case
An illustrative scenario, not a real incident.
A review agent should only read. With its tools limited to read access, it cannot edit, even if a prompt tells it to.
See it in different ways
mindmap
root((Agent limits))
Identity
Write access to trigger
Repository
One branch
Branch protection
Tools
MCP allow list
Tool permissions
Network
Firewall allowlistflowchart TD
Q{"Is the tool allowed?"}
Q -->|"Yes"| A0["Agent can call it"]
Q -->|"No"| A1["Agent cannot call it"]Key point: Tool permissions limit what an agent can do.
Why the other options are wrong
- A. They control how many people can read the agent's session log
- They do not control who reads logs.
- B. They decide which model the agent uses for each of its turns
- That is a model setting, not a tool permission.
- D. They set the budget for premium requests
- They have no link to a request budget.