Guides & sensors

Böckeler's harness framing: GUIDES steer the agent before it acts (feedforward); SENSORS catch mistakes after (feedback) — each either computational (deterministic: linters, tests, type-checkers) or inferential (semantic: AI review, LLM-as-judge).

Connected concepts

Harness engineering, Verification loops, Guardrails / bounded autonomy, LLM-as-judge, Deterministic verification of generated code