The spawnagent.sh blog
Practical notes on running AI agents from the command line: how they work, how to contain them, and how to run many.
-
What It Means to Spawn an AI Agent From the CLI
An AI agent is a model, a loop, tools and context. What spawning one from the command line involves: lifecycle, headless runs, outputs, exit codes and timeouts.
-
Sandboxing Coding Agents: Permissions, Worktrees, Limits
How to contain a coding agent: least privilege, permission modes, git worktrees, containers and VMs, network egress rules, secrets out of reach, and spend caps.
-
One Agent or Many? Running AI Agents in Parallel
When running AI agents in parallel pays off, when it backfires, how to split work, track who owns what, catch merge conflicts early, and check the result.