See the consequential action.
Start with Mara’s F-104 case and separate a lookup, a proposal, and an effect that changes an external resource.
GOV · public course preview
Set the boundary before an agent changes money or access.
For tech leads, product engineers, and operators who can describe an agent, a tool, and an intended business action, this course helps you build an action-boundary record for a new fictional case. You will be able to defend a limited PROMOTE or HOLD decision for a proposed action and, separately, require ROLLBACK as recovery only when an unsafe effect has already executed—without treating the record as production authorization, IAM design, or compliance proof.
This public preview opens GOV.1 and the complete course map. The no-code workbook, offline labs, and capstone remain part of the full course path; enrollment and checkout are not available yet.
A refund assistant may look up Mara’s order F-104 and propose a refund. Sending money to a payment provider is a separate action. The exercises ask what must be bounded, confirmed, retained, and recoverable before that effect can advance.
One decision at a time — with a way to check yourself.
This is not a reading list or a black-box checker. Every module gives you one decision from the same fictional case, asks you to commit before the answer, shows the reason, and carries the result into the next move.
Make a limited call.
Choose the boundary, evidence state, or recovery path before opening the self-check or running a script.
Compare your reason.
Diagrams, worked contrasts, and the answer explain why the record stays HOLD or why recovery becomes ROLLBACK.
Use the next artifact.
The lab checks one bounded slice; the next module and capstone make you reconcile it with the rest of the record.
If you get stuck, the next move is named: return to the case, inspect the contrast, make the small decision, then use the checker’s field-level evidence. The course never asks you to infer a production authorization from a green result.
Before you start
This course is for tech leads, product engineers, and operators who can describe an agent, a tool, and an intended business action. You need a browser and a terminal for the builder path. You do not need advanced OAuth knowledge or legal-audit training.
where GOV stops
GOV makes one tool action reviewable. It does not teach repository governance or threat analysis. A completed record does not grant access, approve a payment, prove compliance, certify production safety, or replace human accountability.
Choose the outcome you need
The starting practice is shared. The difference is what you do after it.
Challenge the action before reading an implementation.
Complete the workbook, then use the six modules’ cases and self-checks to defend the record’s boundary, evidence, decision, and limitation. Running Python is not required for this reviewer outcome.
Build and reconcile the six-part record.
After the workbook, read GOV.1–GOV.6, run each module’s deterministic Python lab, keep its artifact, and assemble the fictional capstone. The six labs are required only for this builder outcome.
Six decisions, one action-boundary record
Each module teaches one review decision. Its linked lab is the corresponding builder exercise; complete all six labs only when you are pursuing the builder outcome.
Name the action
Separate consultation, proposal, and the tool call that changes money.
Lab · action-inventory →Draw authority
Bind the role to F-104, one audience, and a concrete limit.
Lab · authority-boundary →Make confirmation meaningful
Show the precise action and consequence before a person chooses.
Lab · meaningful-confirmation →Keep reviewable evidence
Retain the request, decision context, tool result, and reviewer note.
Lab · reviewable-evidence →Recover after an unsafe action
A blocked request stays on HOLD. An unsafe action already executed needs ROLLBACK.
Lab · unsafe-action-drill →Reconcile the six-part envelope
Choose HOLD for an incomplete proposal and ROLLBACK for an unsafe action already executed.
Lab · release-decision → Next · assemble the capstone →Assemble one fictional action-boundary record.
The capstone composes the six lab artifacts: action inventory, authority boundary, confirmation, evidence package, unsafe-action recovery, and release decision. A passed synthetic structure is not production approval. The final record must keep a pending request on HOLD and identify ROLLBACK only when an unsafe effect has already happened.
What completion demonstrates
For a new fictional refund-assistant case, you can defend a bounded action record and explain why it supports PROMOTE, remains on HOLD, or requires ROLLBACK. The course uses local synthetic practice. It includes no cloud account, payment system, production access, community, certificate, or promise of production safety. Workload measurement is still pending.
Read GOV.1, then inspect the complete path.
The offer includes the full EN/PT course, the required no-code workbook, six offline Python labs, a starter-bundle builder, and the fictional capstone record. It is individual self-study for technical leaders, product engineers, and operators.
offer · USD
US$129
Regular price: US$149
This conservative pre-launch offer is shown on the EN route. The final total shown at checkout is authoritative; local taxes, payment-method costs, or conversion can change it before confirmation.
15-day guaranteeRequest through Hotmart under the rules shown at checkout.
Yours to keepPersonal license for the downloaded release, with 12 months of package updates.
Offline deliveryBilingual ZIP in the Hotmart library after payment confirmation; labs run on a computer.
Focused supportEmail help for access, download, and a reproducible first-run issue.
Individual self-study. No certificate, academic credit, live tutoring, implementation consulting, community, or lifetime updates. A terminal and familiarity with an agent that calls tools are prerequisites; every exercise remains offline and synthetic.