🚀 Execution and operations
The commands in family 07: start with the next possible action, break it into tasks with owners, estimate without false precision, prioritize, parallelize, automate — and review against criteria.
▶️ /execute — turn a plan into action
The command that ends planning. /execute starts with the next possible action and produces concrete artifacts or operational instructions — the email text, a file skeleton, a command to run. It doesn’t describe the work: it does the first part of it.
✓ /execute output
- ✓The email ready to send, with a subject and recipient.
- ✓The configuration file filled in with values from your context.
- ✓"Done: X. Blocked on: Y (needs credential Z). Next: W."
✗ Disguised planning output
- ✗"First, you should prepare the environment, then..." — generic instruction, zero deliverable.
- ✗A list of 14 steps without doing any of them.
- ✗"I recommend aligning expectations with stakeholders."
🎯 Copy and run
/execute a partir de <plano ou blueprint>; comece pela próxima ação possível, produza o artefato concreto (texto, arquivo, comando) e reporte feito / bloqueado / próximo
How to verify: you received something ready to paste or run — not a description of how it was produced.
New here? Artifact is the tangible result of an action: a file, finished text, command, or completed table. "I analyzed the problem" produces no artifact; "here’s the table with the 5 problems and the cause of each" does.
➡️ /next — the next best action
Choose the step for highest expected value considering blockers and reversibility. This is the command for when there are ten possible things to do and paralysis starts to cost more than any wrong choice.
🧭 How /next decides
- •Expected value: how much the step advances the goal × probability of success.
- •Unblock: a step that unlocks others is worth more than an isolated step of equal value.
- •Reversibility: in a tie, the easiest step to undo wins — you learn at lower cost.
🎯 Copy and run
/next dado <situação atual e opções em aberto>; escolha UMA próxima ação com justificativa por valor esperado, desbloqueio e reversibilidade; diga o que NÃO fazer agora
How to verify: the response is one action, not three. And the “not now” list exists — it’s half the value.
💡 Practical tip
Use /next at the end of every work session: "given what I did today, what’s tomorrow’s next action?" Starting the day already knowing the first step eliminates the half-hour warm-up.
🧩 /tasks + /owner — break down and assign
/tasks turns an objective into units small, verifiable, and attributable. /owner defines an owner, support, approver, and dependencies for each one — because a task with two owners has none.
| Role | Question it answers | How many per task |
|---|---|---|
| Owner | Who makes sure it happens? | Exactly 1 |
| Support | Who helps, if called? | 0 a 2 |
| Approver | Who says, “Looks good, you can proceed”? | 0 or 1 |
| Dependency | What or whom is this waiting on? | As many as there are—declared |
One owner per task is a rule, not a preference. "Marketing team" isn’t an owner.
🎯 Copy and run
/tasks /owner para <objetivo ou etapa do blueprint>; tarefas de no máximo meio dia, cada uma com critério de pronto, dono único, apoio, aprovador e dependências
How to verify: no task has "team," "everyone," or "we" as its owner — and none takes more than half a day.
New here? Completion criterion (definition of done) is the sentence that says when the task is finished: “PR approved and merged,” “page live and responding with 200.” Without it, a task stays 90% done forever.
⏱️ /estimate — estimate without false precision
Estimates time, complexity, cost, and uncertainty — the word missing from every estimate that goes wrong. “3 days” is a polite lie; “2 to 5 days, probably 3, because we’ve never integrated with this API” is an estimate.
Range, not a point
Optimistic / probable / pessimistic. The gap between optimistic and pessimistic IS uncertainty — and that’s information.
What pulls toward the pessimistic case
Name the source of uncertainty: an external dependency, new technology, or an open requirement.
What would narrow the range
A 2-hour spike, a question for the vendor, a disposable prototype. Buying information is usually cheaper than making a mistake.
🎯 Copy and run
/estimate para <tarefa ou conjunto de tarefas>; faixa otimista/provável/pessimista, fonte principal da incerteza e a ação de 2 horas que mais reduziria a faixa
How to verify: no estimate is a single number, and every wide range comes with a reason for being wide.
📊 Why ranges work
Whoever asks for “a number” gets the optimistic one and holds you to it as a commitment. Whoever asks for a range gets the uncertainty spelled out and can decide: accept the risk, buy information, or cut scope. The range isn’t imprecise—it’s the precision that’s possible, stated honestly.
🎯 /priority — prioritize by impact and risk
Sorts tasks by impact, urgency, dependencies, risk, and effort. Five axes — because prioritizing only by urgency means letting other people's inboxes decide your day.
Illustrative: family 07 is a ring—/review feeds back into /next, and the ring spins until the milestone lands.
✓ Defensible priority
- ✓"Payment integration first: high impact, unblocks 4 tasks, external risk — start today."
- ✓Short list: 3 priorities. The fourth is "no."
✗ Noise-driven priority
- ✗"First, whatever the client complained about most yesterday" — urgency without impact.
- ✗A list with 11 priorities. Eleven priorities is none.
🎯 Copy and run
/priority sobre <lista de tarefas>; pontue impacto, urgência, dependências, risco e esforço, entregue as 3 primeiras com justificativa e declare o que fica explicitamente para depois
How to verify: the justification for each of the 3 cites at least two dimensions — and the “for later” list isn’t empty.
⚙️ /parallel + /automate — parallelize and automate
/parallel separates what can run at the same time from what requires sequencing. /automate looks for repetitive tasks with clear rules, triggers, and APIs — the candidates to remove from your agenda forever.
🤖 Signs it can be automated
- •Repetition: the task happened 3+ times in the last month, the same way.
- •Clear rule: you can explain the decision as “if X, then Y” without saying “it depends.”
- •Identifiable trigger: an event fires (an email arrives, a spreadsheet changes, the time comes).
- •Available interface: API, webhook, export—something a machine can handle.
🎯 Copy and run
/parallel /automate sobre <lista de tarefas ou SOP>; agrupe o que roda em paralelo vs. em sequência, e para cada tarefa repetitiva diga: regra, gatilho, ferramenta candidata e o que ainda exige humano
How to verify: the “still requires a human” column exists and is honest — automation that pretends there are no exceptions becomes an incident.
⚠️ Automating a mess multiplies the mess
Only automate a process that already runs well manually (Module 4.1, /SOP). Automating a broken process produces the same error, faster, with no one watching.
📡 /monitor — indicators and thresholds
Specifies indicators, frequency, thresholds, and actions when something changes. Monitoring isn’t looking at a dashboard: it’s deciding in advance what will trigger what.
| Indicator | Frequency | Threshold | Action |
|---|---|---|---|
| Landing page conversion rate | Daily | < 1.5% for 3 days | Run /causes on the drop; test a title variant |
| Lead response time | Per event | > 2h | Alert the group; reply with the standard template |
| Automation errors | Hourly | > 0 | Pause the workflow; open a task with an owner |
Teaching example: every row is complete—an indicator without an action is curiosity, not monitoring.
🎯 Copy and run
/monitor para <projeto ou operação>; 3-5 indicadores com frequência de leitura, limiar numérico e a ação exata quando o limiar for cruzado
How to verify: every threshold is a number with a unit, and every action starts with a verb. "Monitor closely" is not an action.
💡 Practical tip
One health metric and one outcome metric are enough to get started. Ten metrics without thresholds become wallpaper; two with defined actions change behavior.
🔁 /review — review against criteria and make corrections
Compare the result with the criteria set earlier, record deviations, lessons learned, and corrections. It’s the command that closes the loop: without it, execution becomes repetition—with it, it becomes improvement.
✓ A review that teaches
- ✓"Goal: 10 sign-ups. Result: 6. Deviation: -40%. Likely cause: link only in the story, not the feed. Fix: pin it to the profile + repost."
- ✓Learning recorded where the next /blueprint will read it.
✗ Feelings-based retrospective
- ✗"It was good, the team worked hard, next time we’ll go even stronger."
- ✗No numbers compared with the original criterion — because the criterion was never written down.
🎯 Copy and run
/review de <entrega ou ciclo>; compare resultado com os critérios e métricas originais, liste desvios com causa provável, 3 aprendizados e as correções que entram no próximo ciclo
How to verify: each deviation has a number and a cause — and at least one fix becomes a task with an owner (/tasks /owner).
💡 Composite recipe
Complete execution recipe, straight from the document: /blueprint /dependencies /priority /parallel /execute /monitor /review. Run it once per cycle, and the next blueprint is already born corrected.
Quick self-check (optional): what's the correct output of /execute?
🎓 Module Summary
Next Track:
5.1 - Product, business, and monetization — you know how to plan and execute; now apply that to the market, customer, offer, and price.