Harness Check Designer

Use this skill to design checks for agent-harness. The harness only runs check commands and stores reports; each check script owns its domain logic.

Positioning

Agent Harness Engineering means engineering the system around an AI agent so the agent's work is constrained, observable, verifiable, and improved through executable feedback loops.

agent-harness is a focused reference framework, not an eval platform, agent runtime, or built-in rule library. This skill is for designing artifact acceptance loops and workflow-controller projects: define artifact contracts, encode project-owned checks, write agent-readable reports, design workflow state routing, and wire blocked checks back into the next agent pass.

Workflow