Agent Concepts vs Linux Concepts
A lot of agent infrastructure is old Linux ideas with new names.
| Agent | Linux |
|---|---|
| Identity | User |
| Least privilege | Permissions |
| Workspace | Home directory |
| Runtime | Process |
| Scheduler | cron |
| Observability | Logs |
| Human in the loop | SSH |
| Isolation | Container / VM |
| Self healing | Restart policy |
| Resource limits | cgroups |
| Loop | for i in $(seq 1 $MAX_ITERATIONS); do; ... done |
| Tool access | Executable permissions |
An agent is an untrusted process that can use tools. A lot of the execution layer concepts already exists in Linux.