agents-best-practices

agents-best-practices is a provider-neutral Agent Skill for designing, auditing, refactoring, and explaining agentic harnesses. It is built around the principle that the model proposes actions, while the harness validates, authorizes, executes, records, and returns observations. The project applies to coding agents, research agents, support agents, operations agents, sales agents, finance agents, healthcare agents, education agents, and workflow automation agents. It helps users reason about tool permissions, runtime discipline, observability, evaluation, and safer execution boundaries. The skill can also generate MVP blueprints for agent systems without tying the design to a single model provider. It is useful for teams building AI agents that need reliable control layers instead of loose prompt-only behavior.