It’s not enough to create agents.
You also need to manage them.
Learn how to delegate work to the AI with a clear goal, the right limits, and supervision that fits your week.
From the 7 principles to the sheet for your first agent: visual explanations, cases from 10 areas, copy-ready examples, and a final project.
A cycle, not a list.
intenção → contexto → dados → critério
→ autonomia → observação → supervisão ↺From turning point to an agent in production.
Follow in order to reach the final project with your agent sheet ready. Already use agents? Enter the path that solves your next challenge.
The turning point
From AI that answers AI that works—and what changes in your role.
From tool to team / Separate marketing from the methodExplore the path →Delegate well
Intention, context, data, and success criteria: what the agent needs to know before starting.
Intention and context / Reliable data and success criteriaExplore the path →Autonomy and supervision
How much freedom to give, how to follow along, and when to level up.
Autonomy with limits / Observation and supervisionExplore the path →The secret and the project
People management under a different name, code agents, and your agent in production.
The secret: you already know this / Final project: your agent in productionExplore the path →Your first agent sheet, tested and supervised.
The final project turns a real task from your work into a well-managed agent: mission, house rules, source of truth, definition of done, calculated autonomy level, report, triggers, and a review ritual. Cases from 10 areas follow your journey, from a clinic to a marketing agency.
- Choose a frequent task, with a visible output and cheap errors.
- Answer the 7 questions and calculate the autonomy level.
- Run the 3 tests before trusting—and adjust one rule at a time.
- Supervise the first week with a failure log.
Open, fill in, use.
All free, no login, running in the browser. The skill is optional for those using Claude Code in the terminal.
Study with your own trail.
Mark topics as read, flag questions, and select excerpts from the explanations to highlight or take notes. In My learning journey, collect your notes, export a backup, and pick up from the last topic.
Progress measures topics marked as read; it is not a certification of mastery. Do the exercises before opening the answers. Your notes stay in the browser and are not sent to a server.
Prerequisites: none technical knowledge required. Just have used ChatGPT, Claude, or Gemini at least once. Every new term is explained the first time it appears.
Sources and editorial criteria
The course starts from three project 7PA texts: a report on the new work logic with AI agents, a critical analysis that separates the promotional layer from what is technically relevant, and a synthesis of the 7 principles. The originals are in github.com/inematds/7pa/conteudo. Examples, exercises, diagrams, and cases were developed for this course.
The autonomy level calculation rule (N0 to N4) and the criteria for leveling up are an educational proposal from the 7PA project, designed to start safely. They are not a technical standard or a market benchmark. Adapt to your context, always in the direction of more control when in doubt.
Mentions of programming agents (Claude Code, Codex, CLAUDE.md, AGENTS.md, permission modes) describe capabilities available in September 2026. Names and options may change depending on the version and the account. Duration estimates include hands-on practice; they are not measurements of students.
Project tools: Agent Sheet, sheet gallery, guide to the 7 principles and The Secret of the 7 Principles.