One honest job: structure Grok completely. Seven layers. Then a first-setup checklist you can actually finish.
01 · Foundation
Global Custom Instructions
Settings → Custom Instructions. Define voice, core principles, default output format, and how uncertainty is handled. Keep this layer light. Heavy domain knowledge lives in Projects.
02 · Personas
Custom Agents
Available on Plus and Heavy — verify live slots. Give each agent a clear name, a focused instruction set, and only the tools it needs. On Heavy, instruct them to collaborate in parallel.
03 · Domains
Projects / Workspaces
One Project per major life or work area. Each can hold its own instructions, files, isolated history, and tools. Global stays modular. Projects carry the weight.
04 · Continuity
Memory
Enable and curate Memory so Grok keeps facts, preferences, ongoing work, and past decisions. Review regularly. Delete anything outdated. Memory is not a junk drawer.
05 · Repeatable work
Skills and Modular Workflows
Define reusable skills as slash commands, saved prompts, or sequences you paste. Every step, check, and format is yours to replace. Illustrative examples are never the final version.
06 · Force
Modes and Tools
Expert / Think for deeper reasoning. On Heavy, spawn parallel specialists and synthesize. Enable DeepSearch, code, image, video, and X/web search per Project — not everywhere by default.
07 · Your voice
Personalization Layer
Refine continuously: preferred phrasing, domain glossaries, decision rules unique to you, and feedback loops. When you say X, it does Y. Structure is living.
First setup
- 01Set Global Custom Instructions with your core rules.
- 02Create 2–4 Custom Agents for your most common roles.
- 03Create one Project for your highest-priority work and load files + instructions.
- 04Turn on / clean Memory.
- 05Write 1–3 Skills for repeated workflows.
- 06Start a fresh chat inside the Project and test.
- 07Adjust based on real output.