A roundup of four topics around AI agent development and operations in February: a study showing AGENTS.md may be counterproductive, Continue.dev's CI-integrated AI checks, AWS Strands Agents' built-in session persistence, and Docker Shell Sandbox for isolated agent execution.