A year ago these were engineering words. Now they're how AI products get built. ---- Hooks = rules enforced by code Tools = what the agent can reach Skills = playbooks it loads on demand Memory = what it still knows tomorrow Harness = everything around the model Sandbox = a walled-off place to run code Permissions = what it can do without asking Context Library = the folder it reads first Rules = standing instructions saved as files Checks = what catches mistakes before you do Instructions…