The RoboNuggets Community

Agentic AI Foundations

Multiple Modules. 50+ Topics. Tool-agnostic. The mental model behind every agent — so whether it's Claude Code, OpenClaw, Hermes, Antigravity, or whatever comes next, it just clicks.

Multiple modules· 50+ unlocks· Community-exclusive
M1
M1 · ANATOMY

The Anatomy of an AI Agent

Tools change weekly. The mental model behind every AI agent doesn't. Get this right once and every new launch clicks.

1.1The Mental Model
  • Pace of change
  • The durable pattern
1.2Chatbots vs Agents
  • How chatbots work
  • How agents differ
1.3The Model
  • Inside an LLM
  • Opus · GPT · Gemini
1.4The Harness
  • Inside a harness
  • Claude Code · Codex · Antigravity
1.5The Context
  • Plain English instructions
  • CLAUDE.md and AGENT.md
1.6Agentic Loop
  • Look · think · act · check · repeat
  • Stop conditions
M2
M2 · DEEP DIVE

The Model, the Harness, and Context

Going deeper on the three parts. What each one actually does — and how they fit together.

2.1The 6-Month Rule
  • Plan for next-gen models
  • Spikiness explained
2.2Model Tiers
  • High · mid · low
  • Anthropic · OpenAI · Google
2.3Tokens
  • Tokens explained
  • Bible benchmark
2.4Context Window
  • Working memory
  • Overflow problem
2.5Harness Surfaces
  • Terminal · desktop · IDE · messaging
  • IDE = sweet spot
2.6CLAUDE.md and AGENT.md
  • The agent's job description
  • Pre-injected every session
2.7Skills 101
  • Inside a skill
  • One file, one workflow
2.8Memory 101
  • How memory works
  • MEMORY.md index pattern
M3
M3 · LANGUAGE

The Language of Agents

Demystify the jargon. Every term you'll see in docs, tweets, and tutorials — in plain english.

3.1IDEs
  • IDEs explained
  • VS Code · Cursor · Antigravity
3.2Permission Modes
  • Auto · Ask · Plan · Bypass
  • When to use each
3.3Tool Permissions
  • Per-project tool rules
  • Allow · ask · deny
3.4Effort
  • How effort works
  • Token implications
3.5Model Picker
  • Opus · Sonnet · Haiku
  • When to switch
3.6Model Usage
  • Token spend tracking
  • Quotas and rate limits
3.7Context Hygiene
  • Token bloat problem
  • Pruning techniques
3.8Compact & Clear
  • What /compact does
  • What /clear does
3.9Skill Patterns
  • Trigger conventions
  • Frontmatter + body
3.10Memory Layers
  • Index file
  • Individual files
  • External systems
3.11MCP Servers
  • MCP explained
  • Gmail · Drive · Slack · Linear
3.12CLI Tools
  • Bash · git · gh
  • Built-in tool calls
3.13Plugins
  • How plugins work
  • Custom slash commands
3.14Routines
  • How routines run
  • Cron syntax basics
3.15Hooks
  • How hooks fire
  • Pre · post triggers
M4
M4 · SKILLS & ORCHESTRATION

Core Agentic and Orchestration Skills

The skills that turn one agent into a team. Build your own, calibrate as you go, run multiple sessions on one project.

4.1Build your Skills
  • Skill anatomy
  • master-* prefix convention
4.2Calibrate
  • End-of-session ritual
  • Light · full · auto modes
4.3Coordinate
  • Shared project folder
  • context.md + session-log
4.4Onboard
  • Pick up any project
  • Brief before effort
4.5Align
  • Clarifying questions first
  • Scope before effort
4.6Devil
  • Contrarian view
  • Break the yes-man pattern
4.7Burst
  • N variations on demand
  • Pick the winner
4.8Tweak
  • Live UI sliders
  • Bake the chosen design
4.9Session Continuity
  • Convo logs · handoffs
  • Fork conversation
4.10Multi-Agent Orchestration
  • When to split
  • Why specialists win
M5
M5 · COMMAND CENTRE

Your AI Command Centre

Every agent. Full visibility. The dashboard layer nobody else teaches.

5.1The Why
  • Visibility at scale
  • The 3-agent threshold
5.2The Dashboard
  • Single-screen overview
  • Live agent state
5.3Agent Statuses
  • Idle · working · blocked
  • Live polling
5.4Sprint & Backlog
  • The sprint board
  • The backlog queue
5.5Generations
  • Image · video · audio
  • Cost per generation
5.6Docs
  • Living markdown docs
  • Agent read + write
5.7Flows
  • Multi-step pipelines
  • Schedule or trigger
5.8Skill Tree
  • Skill map per agent
  • Spotting duplication
5.9Building your Own
  • Open-source starter
  • Or fork ours
ROBOLABS