Your agent says it is done. Now you can check.
A 13 module system for writing work an agent can execute, delegating it safely, bounding the loops it runs in, finishing against checks you wrote before the work started, and packaging what worked as a skill you run again.
13 modules, 39 lessons About 8.9 hours of material
Tax included
Payment is handled by Stripe.
After this course
After this course you can hand a real job to an agent and verify what comes back. You turn a loose request into an eight section work order, run a planning session whose only output is a document, and keep the project's memory in four files so every session opens by reading the record. You split work across agents with separate silos, stand an independent reviewer in front of anything consequential, bound every loop so one going nowhere stops on its own, and close against checks written before the first line of work. You strip the client names out of the workflow that worked and package it as a skill you can run again, and an agent operating a logged in tool leaves every object paused or in draft. By the end, done is something you looked at yourself.
What you will be able to do
Turn a voice note into an eight section work order: Objective, Background, Instructions, Done when, Watch out, Systems and paths, Do not, Defaults chosen.
Run a planning session with Claude Fable 5 that returns work packages carrying an owner, verified inputs, bounded scope, a complete when sentence and a test.
Keep discovery.md, research.md, plan.md and progress.md so any session, yours or an agent's, opens by reading the project record.
Stand an independent reviewer in front of consequential work with two verdicts, PASS or BLOCK, where a BLOCK holds until the artifact is repaired and reviewed again.
Bound a loop with iterations, time, scope, authority and written blocker conditions, plus a completion promise a stranger could grade.
Scope a workflow you have already run into one reusable skill, strip the client names and credentials out of it, then test it blind the way a stranger would run it.
Who this is for
- The solo marketer whose AI output looks impressive and cannot be proven finished.
- The freelancer who wants agents to carry delivery and keeps rewriting whatever they hand back.
- The Claude Code or Codex user who reaches eighty percent and then watches the run wander.
- The operator whose best workflow lives inside one long chat session and has to be rebuilt from memory every time it is needed again.
Who should skip it
- You want a prompt library to paste from. This course spends its hours on work orders, checks and evidence, and it asks you to read what the agent produced.
- You want speed with no paperwork. A work order, a plan document, a project record and a written check all come before the agent starts, and that front loading is the method.
Inside the course
Every module and every lesson, exactly as they sit in the course today.
The Full Agentic Workflow5 lessons
Module 1
The six stages from voice note to finished system, the four roles that own them, and your first five skills installed and inspected.
- From Voice Brain Dump to Finished System12 min
- Planner, Implementer, Operator, and Validator12 min
- Define Done as Observable Behavior1 min
- Choose One Agent, a Swarm, or a Loop12 min
- Install the Agentic Engineering Skill Pack13 min
Prompt Engineer as a Work Order Compiler3 lessons
Module 2
Ten minutes of context gathering, the eight section work order, and permission that survives a twenty minute unsupervised run.
- Gather Context Before You Transform13 min
- Write the Eight Section Transform13 min
- Preapproval Without Authority Drift13 min
Planning With Claude Fable 53 lessons
Module 3
A session whose only output is a document, feasibility settled with evidence, and a plan cut into work packages with fallbacks.
- Use Fable as the Planning Orchestrator12 min
- Research the Current System and Feasibility13 min
- Produce an Executable Plan and Fallbacks16 min
Harness Engineering3 lessons
Module 4
Four files that hold the project's memory, plus breadcrumbs and a source hierarchy for the day two sources disagree.
- Build the Four File Harness1 min
- Read Progress First and Log Failures14 min
- Breadcrumbs and Source Hierarchy15 min
Swarm Design and Delegation3 lessons
Module 5
Orchestrators assign and workers own silos, with output contracts and one real item pushed all the way through first.
- Orchestrators Plan and Workers Own Silos1 min
- Parallel Work, Pipelines, and Shared State15 min
- Structured Contracts and the First Complete Path14 min
Implementation With Codex3 lessons
Module 6
Enter the code through the plan, keep changes small, run workers in separate folders, and require evidence with every finished piece.
- Enter the Repository Through the Plan15 min
- Minimal Changes, Worktrees, and Supervision15 min
- Integrate and Return Evidence14 min
Playwriter Browser Operations3 lessons
Module 7
Observe before clicking, work in draft and paused states, read every change back, and keep an audit trail of the session.
- Connect, Observe, and Identify the Target14 min
- Operate Marketing Tools Through Draft States14 min
- Readback, Recovery, and Audit Trails15 min
Guardrail Agents and Approval Gates3 lessons
Module 8
A second agent that only judges, five guardrail patterns matched to named failures, and approvals that name exactly one action.
- Design an Independent Reviewer14 min
- Use the Right Guardrail Pattern15 min
- Approve the Exact Action and Read It Back15 min
Validation Is the Definition of Done3 lessons
Module 9
Checks chosen from your finish line, runtime and persistence testing, then repair, revalidation, and a report anyone can verify.
- Choose Project Aware Checks14 min
- Test Runtime Behavior and Persistence15 min
- Repair, Revalidate, and Report Evidence17 min
Loop Engineering3 lessons
Module 10
Five loop types, a completion promise a stranger can grade, and five bounds so a loop going nowhere stops on its own.
- Choose the Loop and Completion Promise17 min
- Run Repair, Polling, and Lifecycle Loops18 min
- Bound the Loop and Preserve Learning20 min
Package Workflows as Public Skills3 lessons
Module 11
Scope a skill to one job, strip client names and credentials, test it blind the way a stranger would run it, and version it.
- Design a Focused Reusable Skill18 min
- Convert Private Workflows Into Safe Packages19 min
- Validate, Test, and Version the Skill19 min
End to End Capstone3 lessons
Module 12
Design one real system, run it on real inputs, repair from evidence, and package it so the next version starts from this one.
- Design the Capstone System15 min
- Run the System and Repair With Evidence14 min
- Demonstrate, Package, and Compound1 min
Downloads: Skills, Agents and Workbooks1 lesson
Module 13
The 21 installable skills and every module workbook, with how to read a SKILL.md and its guardrails before the first run.
- Everything in This Course, In One Place24 min
What comes with it
Every lesson in the course, with its written material and its downloads.
One payment. The course stays on your account and opens on any device.
A certificate once you finish it.
Five agents from the library of 82, open with any purchase on this store.
Questions
- Why buy this alone for $99 when the bundle is $497?
- One course is $99 and you own it. All 16 bought separately come to $1,584, so the $497 bundle wins the moment you want three or more. Take this one on its own if agentic engineering is the piece you need now.
- Do I need to be technical?
- Module 1 assumes nothing. No tools are installed yet, no coding is involved, and the first exercise is a voice note about a process you already do by hand. What the course does ask is that you read what an agent produced and check it.
- What software do I need, and does it cost extra?
- An AI tool that loads skills and can read and write files on your computer. Claude Code and Codex both do. Playwriter borrows the browser session you are already logged into, so credentials stay out of prompts and files. Those tool accounts are billed by their own providers and sit outside this purchase.
- Are the agents included?
- The course downloads come with the course: 21 installable skills, a workbook for every module, and the Playwriter prompt set. The separate agent library on this platform holds 82 agents, any purchase unlocks 5 of them, and the remaining 77 stay visible with a price or arrive with the Skool community.
- How long does it take?
- 13 modules, 39 lessons, about 8.9 hours of material. Module 12 runs longer than its share of that, because it asks you to build and run one real system rather than read about one.
- Can I get a refund?
- Access is granted the moment payment completes, and this is a digital product, so a purchase is final once access has been given. The refund page states it in full.
Start today
Agentic Engineering
13 modules, 39 lessons
Or take all 16 courses
You save $1,087, 69% off
16 courses, 151 modules, 532 lessons
Three courses bought separately already cost more than half the bundle. The bundle carries all sixteen for one payment.