Advanced Claude Code: Context Engineering, Skills, and Workflows for Personal Automation
Claude Code is not just for coding. It is the most powerful AI agent you can run on your own computer, and it can do almost anything you can, from reading and editing files to installing apps, running code, and browsing the web. To put that power to meaningful work, master three things, context engineering, skills, and agentic workflow patterns, and apply them to everything you do.
Designed for non-techies, taught by a technopreneur and 19-year industry veteran
Most people meet Claude Code as a coding tool. It is something much bigger: the most powerful AI agent you can run on your own computer, one that can do almost anything you do at a keyboard, from your inbox and reports to your meetings and client work. This workshop is for non-techies who want to use it that way: consultants, managers, analysts, ops people, and founders who do not code. If you can install an app and follow typed instructions, and you are willing to treat the terminal as a place to chat with Claude, you are a good fit.
This is not a coding course, and it is not a feature tour. Out of the box Claude Code is generic; what turns it into something that works the way you do is mastery of three things: engineering its context, packaging your knowledge as skills, and composing those skills into agentic workflows. You learn each one in plain English, then practise it in a hands-on activity designed to make the concept stick, so you leave having actually done it, not just watched a demo.
You learn it directly from Boon Kgim, a technopreneur and builder who runs his business on AI and uses Claude Code daily for far more than code. You get the shortcuts, the patterns, and the gotchas in one day, instead of working them out yourself over months. Making complex things simple for people who do not code is how he teaches.
Why learn from Boon Kgim
- A technopreneur who powers his whole business with AI, with roots as a tech builder, consultant, and growth marketer
- Built 10+ apps with Claude Code and developed a battle-tested workflow
- Uses Claude Code daily beyond coding: presentations, proposals, workshops, automation, marketing, and more
- A track record of teaching non-technical professionals who walked in nervous and walked out with real, working results
- Direct access during the workshop, you ask, he answers, no TA layer in between
- Access to the alumni community group chat, plus free re-attendance to a future run
You are not paying for a course. You are paying for the compressed experience of 19 years building software and automating real work, with Boon Kgim sitting next to you while you learn to make Claude Code work the way you do.
What actually unlocks Claude Code
Out of the box, Claude Code is generic, and most people never get past that. What separates a power user is mastery of three things, and this workshop is built around them in order.
Context engineering comes first, because context is the single biggest lever on output quality. The same model on the same request gives a generic answer with the wrong context and expert-level work with the right one. Every building block you can add, a CLAUDE.md, a skill, a subagent, an MCP connector, the files you point it at, spends that window differently, so context engineering is the skill of choosing the right one for the job and keeping out the noise that drags quality down. CLAUDE.md is simply where you start, the simplest of them.
Skills are the core of the workshop. A skill packages a piece of your knowledge or a recurring task so Claude runs it your way every time. You build across four levels, instruction, templates, scripts, and live systems through MCP, on the rule climb only as high as the job needs, which turns “I don’t know how to code” from a barrier into a non-issue, because most high-value skills live at the first two levels.
Agentic workflows come last, where subagents and patterns like route, chain, and fan-out compose your separate skills into systems that run a whole multi-step job, not just one step. This is loop engineering, the skill of designing the act-check-refine cycle the agent runs to correct its own work. From here that loop can even run without you, on a schedule or when something happens, which is what earns the word automation.
You learn all three through carefully curated, hands-on activities, so by the end of the workshop you have not just understood them, you have used them, building a CLAUDE.md, a skill at each level, a composed workflow, and a capstone automation along the way, plus the mental model to keep building on your own.
You’re not alone. Fall behind? You’re covered
You’re not on your own after the workshop. You get the alumni community group chat, where you share what you build, bring your questions back, and get answers from Boon, who regularly shares his own tips and practical insights and amplifies what members make. And falling behind is never a problem: if you don’t keep up, or you just want a second pass, you can re-attend a future run of the same workshop for free, whenever you want. The seat you take today is one you can always come back to.
Unlock the power of Claude Code with context engineering, skills, and agentic workflows
Claude Code is not just for coding. It is the most powerful AI agent you can run on your own computer, and it can do almost anything you can: read and edit files, install and run apps, write and run code, even browse the web and use it like you would. To put that power to meaningful work, you need to master three things: engineering its context, packaging your knowledge as skills, and composing those skills into agentic workflows. This workshop teaches each building block in plain English, then has you practise it in a hands-on activity, carefully curated to amplify the learning, so you leave having actually done it, not just seen it.
Context Engineering
Why context is the single biggest lever on output quality, and how to manage your window like a shared notebook.
CLAUDE.md
The simplest building block, a plain file Claude reads first every session so it stops asking you the same things.
Skills
Packaging a recurring task once so Claude runs it the same way every time after, loaded only when needed.
The Four Levels
Instruction, templates, scripts, and live systems, the ladder you climb only as high as the job needs.
Templates
Locking the shape of an output so the structure stays consistent every time while the content varies.
Scripts
Automating the deterministic step that must be exact every time, written for you even if you do not code.
MCP
Plugging in ready-made connectors so a skill can draft into Gmail or read your calendar on live data.
Subagents
The composition unit that runs a big multi-step job in its own window without bloating your main session.
Workflow Patterns
Route, chain, fan-out, generate-and-critique, and orchestrator, the shapes that turn single skills into systems.
Loop Engineering
The skill of designing the act, check, and refine cycle the agent runs, so Claude works toward a goal and corrects its own output instead of you prompting every step.
Run Without You
Headless runs, scheduling, and hooks, the step that makes a skill act on its own and earns the word automation.
Automating Safely
Secrets in the environment, idempotency, and planning for failure, so you can trust an automation to run unwatched.
You will walk out with
- A working command of context engineering, skills, and agentic workflows, the three levers that unlock Claude Code
- Hands-on practice writing a CLAUDE.md, the file that personalises Claude Code to a role, voice, and recurring tasks
- A skill built at each of the four levels, instruction, template, script, and live-system action
- Practice composing skills into a workflow that triages and acts
- A capstone exercise that takes one automation end to end
- A clear mental model for context, skills, and when to climb each level, transferable to anything you point Claude Code at next
- The confidence to keep building skills and workflows on your own after the workshop
What you will do in the room
- Run a live context autopsy on a session, then tighten it with the management commands
- Author a CLAUDE.md and feel the before and after difference it makes
- Build and save a Level 1 skill that writes in a consistent voice for a recurring task
- Add a template to a skill so a repeated document keeps its shape with zero drift
- Have Claude write a script into a skill for a step that must be exact every time, then verify it runs reliably
- Connect a skill to a live system through MCP and watch it draft straight into a real inbox
- Compose a router that reads an incoming message, decides its type, and hands it to the right reply skill
Who this is for
- Consultants, managers, analysts, ops, and founders who want to master Claude Code as a general-purpose agent, not just a coding tool
- Non-techies who can install an app and follow typed instructions, and are willing to treat the terminal as a place to chat with Claude
- People who have used Claude Code to build or write and now want to push it to do far more of their actual work
- Software engineers who want to take Claude Code beyond their coding workflow and automate the rest of their work
- Anyone who senses Claude Code can do almost anything they do at a keyboard, and wants the context, skills, and workflow patterns to make that real
Who this is NOT for
- Anyone wanting a developer deep-dive on coding productivity. Software engineers are welcome, but the coding workflow is a separate track
- People who have never opened Claude Code. Start with CC-1, the foundations course
- Anyone wanting a feature tour rather than the deeper craft of context, skills, and workflows, practised through hands-on activities
Indicative pricing
Final quote confirmed on inquiry. Typical reply within 2 business days.
Individual seat
Public cohort registration
- Early birdS$488 per seatLimited-time launch price, usually S$588
Group / Private session
Book a dedicated session for your team
- Up to 8 paxS$4,704 flat8-pax minimum billing
- 9–11 paxS$588 per pax
- 12–15 paxS$548 per paxSave ~7%
- 16+ paxS$508 per paxSave ~14%
Customised session
Full-day default · Also runs well as two 4-hour sessions, splitting after Level 2, when a single day is hard to schedule
What we provide
- Pre-workshop setup guide covering install, sign-in, and connecting your first MCP server, written for non-techies
- A prompt and command cheat sheet for every build, conversational and ready to adapt
- A reference card covering the mental models, the building-blocks scorecard, the four skill levels, the workflow patterns, and the safety checklist
- Access to the alumni community group chat, where Boon answers questions and shares his tips and practical insights
- Free re-attendance to a future run of the same workshop
Upcoming dates
- 9 July 2026 · 10:00am – 6:00pmin-person · Hashmeta Office, 30A Kallang Place, #11-08/09, Singapore 339213
Your facilitator
Boon Kgim
Founder, nanogent.ai
Practitioner at the intersection of AI agents, vibe coding, and growth marketing. Boon teaches from the build, sharing what actually works when shipping AI products in competitive markets. His workshops combine live demos, real workflows, and frameworks you can apply the same day.
Common questions
Do I need to know how to code? +
No. The whole premise is that Claude Code is far more than a coding tool. When the workshop uses a script, Claude writes it for you, and your job is to describe the rule in plain English and verify the result. The terminal is treated as a place to chat with Claude, never a barrier.
Do I need a Claude subscription? +
Yes, you will need a Claude subscription (not included in the course fee). You will use Claude Code signed in to your own account throughout the workshop. Claude Max is recommended for the smoothest experience during and after the workshop.
What do I need set up before the workshop? +
Claude Code installed and signed in, at least one MCP server connected (Gmail and/or Google Calendar) for the live-system activity, and, if you like, a real task of your own to try the activities on. A setup guide is sent ahead of time, and there is a help slot before we start. Doing the install as pre-work is what makes the workshop calm instead of rushed.
Is this hands-on? +
Yes. Every concept is followed by a hands-on activity, carefully curated to amplify the learning, so you leave having done it, not just seen it. Where an activity allows, you can try it on a task of your own. A few people demo their work to the group.
What if I don't finish the capstone in the workshop? +
That is completely fine. People work at different paces, and the goal is the understanding and the habit, not a finished automation by 6pm. You leave knowing how to keep building, with the alumni community group chat to help. You can also re-attend a future run for free to pick up where you left off.
How is this different from the CC-1 foundations workshop? +
CC-1 is the entry point, where you learn the foundations of Claude Code by building your own website. CC-3 assumes those foundations and goes deep on what actually separates power users, engineering context, packaging skills across four levels, and composing agentic workflows, practised through carefully curated hands-on activities. CC-1 makes the basics stick. CC-3 is where Claude Code becomes the most capable agent on your machine.
Is this a feature tour of Claude Code? +
No. Every Claude Code primitive, CLAUDE.md, skills, MCP, subagents, hooks, is introduced as the means to real capability, never for its own sake. You leave having mastered context, skills, and workflows, not having toured a feature list.
From past workshops & talks
Moments from sessions, forums, and stages over the past few years.
Tap to expand