Five Functions of a Harness

The five functions of a Harness form a closed loop: context (sufficiency of information), tools (clarity of interface), constraints (safety by default on failure), verification (input isolation: look at structured data, not freeform model text), and correction (don’t show intermediate errors while recovery is possible). Context and tools enable “can do”, the other three enable “can’t make mistakes”.

Related: [Harness engineering], [Sidecar mechanism], [Failures and recovery]