Judgment-Encoding Hooks

Description

Hook-design philosophy articulated by Blake Crosley in which the hooks that reliably improve agent output quality are those encoding constraints and decision logic (e.g. blocking OpenAI product suggestions inside an Anthropic-only workflow) rather than triggering task automation. Crosley's 84-hook production system is predominantly judgment-encoding; this mirrors the leaked source where stop-hooks are primarily constraint mechanisms (anti-hallucination guards, permission gates, blast-radius checks) rather than workflow triggers.

Key claims

Relations

Sources

src-20260409-e32facfa1c9f