A Week of Grok Bot Lessons in 10 Mins

Summary of technological concepts / product features (GrokBot “hacks”)

This video is a fast guide to improving how you use GrokBot with AI agent workflows, focusing on non-technical setup via plugins/skills, agent delegation, memory control, integrations, logging, training-by-demonstration, and automation routines.

---

9 favorite GrokBot hacks / tutorials

1) Create a “Grill Me skill” (interview-based agent setup)

• Goal: capture the user’s business knowledge/goals so agents “know you.”
• Approach: a “Grill me” skill repeatedly interviews the user until a shared understanding is reached.
• Workflow shown:
  1. Drag the skill into GrokBot
  2. Command it to read/create the skill
  3. Have the agent interview the user
  4. Save the resulting skill
• Usage: later invoke it with prompts like “Hey, grill me about X, Y, Z.”
• Where to find related items: plugins → yours, described as private plugins/skills.

---

2) Set up a “Chief of Staff” agent that delegates

• Pattern: you talk only to one main agent (named Klaus), while specialists handle tasks.
• Delegation rule (from the description): before doing a task, Klaus should check whether another GrokBot “owns it” and delegate first; only do work directly if no specialist fits.
• Specialists shown by role (examples):
  • motion/animation
  • research
  • planning
  • Slack/media work
  • content strategy
• Demonstration: Klaus sends messages to other agents when needed.

---

3) Let “Grill me” decide which sub-agents to build

• If you’re unsure what to specialize, use the Grill Me skill + chief of staff to elicit needs.
• The chief of staff evaluates existing agent coverage and suggests adding higher-value specialists (example suggestions mentioned: tube/video, community, voice) rather than adding redundant general agents.

---

4) Understand GrokBot memory: individual vs shared context

A key concept: GrokBot has automatic memory, but it distinguishes between:

• Individual agent memory (private to an agent/conversation context)
• Shared memory (usable across agents)

Plugin connectivity example: when connected to apps (e.g., Gmail/calendar), agents can share access appropriately—so the shared context should be curated.

The speaker emphasizes specifying:

• what should be shared (e.g., company/channel/funnel/tools)
• what should remain Klaus-only/private

Bonus memory workflow: start “threads” You can select a key message → “start a thread” to have a separate focused conversation about one item, then return to the main session.

---

5) Integrate Compozio to connect to many external apps

• The guide highlights GrokBot plugin setup simplicity.
• If GrokBot doesn’t directly support a specific app (examples: YouTube, Reddit, LinkedIn, GoHighLevel), use Compozio.
• Compozio provides broad “hundreds” of integrations, and GrokBot can connect to Compozio to access those services.
• Shared knowledge example: agents are told Compozio is the route to YouTube/Perplexity/LinkedIn (and future apps).

---

6) Agent logging to an external task system (ClickUp)

• Adds observability: create a skill like “log Grokbot work to ClickUp.”
• Workflow shown:
  • Klaus delegates research to a specialist agent (e.g., Eyes).
  • Results/status are written into a ClickUp space as projects/tasks with:
    • links
    • start time
    • status
    • owner
    • progress updates / notes
• Purpose: track agent work over time and see what’s waiting on whom.

---

7) Teach agents tasks via “learn from demonstration” (screen recording)

• Feature: record your own actions on the agent computer (shared underlying environment).
• Process:
  1. start recording
  2. perform steps (example given: Google Images search/saving results; schedule a weekly rundown, etc.)
  3. stop recording
  4. GrokBot analyzes and converts behavior into a reusable skill
• Framed as easier than explaining complex/visual workflows.

---

8) Create scheduled “routines” for recurring automations

• Routines are recurring scheduled tasks that run on a schedule or triggers.
• You can define:
  • routine name
  • instructions
  • which skills/agents to use
• Triggers mentioned:
  • schedules (hourly/daily/intervals)
  • event-based triggers (examples named: Slack messages, Git events, Teams messages)
• Key benefit: routines run in Grok’s cloud, so they can continue even if your phone/desktop app is off.

---

9) Use the “agent computer” with saved credentials / profiles

• Save login profiles so you don’t paste credentials into chats.
• Example: teach Klaus in a school account that is already signed in → agent can perform tasks in that account (example: 7-day challenge posts: find posts and like them).
• These capabilities can then be automated with routines.

---

Main speakers / sources

• Primary speaker: “Nate” (referred to directly as “Nate” in the chief-of-staff description and throughout the demo context).
• Referenced agents:
  • Klaus (chief of staff)
  • specialist agents like:
    • Motion (motion graphics)
    • Eyes (researcher)
    • Coffee (morning planner)
    • Money (Slack/media)
    • Views (content strategy)