Exam · Orchestrating multiple agents with explicit responsibilities

Take the exam and earn your certificate.

A real, server-scored exam for this course. Score strictly above 80% to pass and get a genuine, verifiable certificate.

Before you start

We ask for your name and email first, the same as a PDF download, so we know you are here and your certificate can carry the right name.

We privately record your name, email and request IP for this exam request. Read the privacy notice and retention policy.

Visit Mickai on LinkedIn

Complete the security check before submitting.

Cloudflare Turnstile helps prevent automated submissions. How this check uses data.

You can retake the exam any time. Each attempt draws a fresh set of questions.

Written practice, ungraded

Write your response in your own notes, then compare it with the sample answer. This practice is separate from the scored exam and does not affect your certificate.

Practice 1

Explain when you would keep a single-function baseline for the fictional Orchard sentence task instead of splitting it into research, draft and review roles, and what you would need to name before splitting.

Compare with a sample answer

I would keep the single function unless I can name a benefit to measure, such as different source access, a specialist quality check, an independent calculation or separate permissions. The baseline has no handoff to corrupt and no reviewer call to pay for. If every worker would see the same context and repeat the same unsupported assertion, the extra roles add calls and failure modes without adding useful evidence.

A good answer names the baseline's advantages, the kinds of benefit that justify a split and the risk of shared context.

Practice 2

Take the valid plan in which write depends on find and check depends on write. Predict the outcome when find is made to depend on check, when write's dependency is replaced with an absent task ID, and when the valid tuple is supplied in reverse order.

Compare with a sample answer

Making find depend on check creates a cycle, and replacing write's dependency with an absent ID creates an unknown dependency; order_plan raises a ContractError for each before any worker is called. The reversed valid tuple still executes find, write, check, because dependency order takes precedence over tuple order. Only the relative order of independent roots can change with insertion order.

A good answer separates preflight rejection of invalid plans from the ordering of a valid reversed plan.

Practice 3

A reply for task find, attempt 1, has the body "Orchard opens at midnight" and refs [s1], while the fictional source s1 says the lab opens at 09:00. Explain which checks the reply passes, why it is still unsupported, and what a task-specific acceptance check and an uncertainty outcome could look like.

Compare with a sample answer

The envelope can pass the task, attempt, field, size and known-reference checks, because s1 is an allowed source for a root task. It still conflicts with s1, which gives 09:00. A fixture-specific check could require the exact recorded opening time; an open-ended reviewer should compare the claim with the source content and report unsupported or uncertain when evidence is insufficient. Known-reference membership alone is not factual validation.

A good answer distinguishes a structurally valid envelope from a supported claim and proposes a task-specific check.

Practice 4

Trace the Orchard demo with a budget of 4 and then with a budget of 2, using retries 1 in both runs. List each invocation with its attempt number, the final state of each task and the accepted results that remain.

Compare with a sample answer

With budget 4 the invocations are find:1, write:1, write:2 and check:1; write:1 returns non-JSON and is recorded as invalid-reply, and all three tasks succeed with three accepted results. With budget 2 only find:1 and write:1 are invoked. Find succeeds, write becomes budget_exhausted before attempt 2 is called, and check is blocked. Only the find result is retained, and the malformed draft reply still consumed a call.

A good answer accounts for every call, including the malformed draft reply, and distinguishes an unattempted retry from a failed one.

Practice 5

In a disposable copy of the three files, the condition calls >= budget is replaced with False. Describe what happens to the budget-2 demo and to the test suite, and explain why the per-task retry cap and the shared budget must be tested separately.

Compare with a sample answer

The modified scheduler ignores the global stop condition, so the budget-2 demo makes four calls: find once, the draft twice and review once. The shared-budget test fails. Restoring the check returns the suite to fourteen passing tests. The local retry cap still limits attempts per task, so it cannot reveal a missing run-wide limit, which is why the two limits need separate tests.

A good answer traces the extra calls, names the failing test and explains why a local cap cannot reveal a missing run-wide limit.

Practice 6

Draft a short release note for the scripted Orchard workflow. Name two checked contracts and the tests that support them, record the observed call counts, and name missing integration properties with the roles responsible for them.

Compare with a sample answer

The reply-identity and shared-budget tests support task-and-attempt binding and the run-wide invocation limit. The successful demo uses four calls; with budget 2 only the research result is retained, the draft is budget_exhausted and review is blocked. Adapter owners must supply deadlines, the storage owner must design durable state transitions, and the evaluation owner must measure factual quality against real sources. No model-quality, concurrency or production recovery claim is made.

A good answer ties each claim to a test or observed output and states what is not claimed.

About this exam

Is this exam graded?

Yes. Every question is scored on our server, never in your browser, and you need a score strictly above 80% to pass. The questions are a real, reviewed bank for this course, drawn fresh for each attempt.

Is it only multiple choice?

Yes. This is a multiple-choice knowledge assessment, with at least one question from each chapter in every attempt. Written exercises are separate practice and are not automatically scored. Passing does not establish practical professional competence or independently verify identity.

Is there a time limit?

Yes, 45 minutes, shown as a countdown once you start. If time runs out, whatever you have answered is submitted automatically.

Do I need to download the PDF first?

No. This course's full text is on its own page, so you can read it online and come straight here when you feel ready.

What do you do with my name and email?

We privately record your name, email, request IP, course and privacy acknowledgement for this request. Newsletter consent is separate and optional. See /privacy.html for storage and retention. A passed exam creates a server-stored named certificate. An email confirmation is not identity assurance.

Can I retake it if I fail?

Yes, straight away. Each attempt draws a fresh set of questions and shuffles the answer order, so it will not be identical to your last attempt, and an attempt can only be submitted once.

Is the certificate real?

It is a genuine, server-issued certificate with its own verification page and a unique ID, not a self-declared one. It records that you passed this course's exam on this date. It is not an accredited qualification, a professional licence or proof of employment.