Quizzes · Learn Kiro

Two rules make quizzes useful instead of decorative: answer before you reveal (commit, even if unsure; the miss is where learning happens), and chase every wrong answer back to its chapter (the revision page collects all chapter summaries for exactly this).

Beginner quiz

Thirteen questions on the ground floor: what Kiro is, how specs, steering, hooks, and MCP fit together, and how the chat modes differ. Read the explanations even when you score the point; they say why the answer is right.

Question 1 · multiple choice

What is Kiro?
Question 2 · multiple choice

Kiro ships on three surfaces. Which list is correct?
Question 3 · short answer

Who builds Kiro, and roughly when did it first appear?

Show model answer

Kiro is built by AWS (Amazon). It launched in public preview in July 2025 and reached general availability in November 2025.

Knowing the source matters: Kiro sits in the AWS family, which is why AWS Builder ID is a sign-in option and why the Amazon Q Developer CLI was later folded into Kiro CLI.
Question 4 · multiple choice

When do Kiro's own docs say a spec session beats a vibe chat session?
Question 5 · multiple choice

A feature spec produces three markdown files. Which one holds the step-by-step plan Kiro actually executes?
Question 6 · short answer

Name the three files a feature spec generates, and say in one phrase what each holds.

Show model answer

requirements.md holds what to build, written as testable acceptance criteria; design.md holds the technical approach; tasks.md holds the broken-down implementation plan.

The three files mirror the workflow phases (Requirements, then Design, then Tasks), so each phase leaves a reviewable artifact behind.
Question 7 · scenario

You keep re-explaining your team's naming rules, folder layout, and testing conventions at the start of every session. Kiro follows them perfectly until the next session, when you have to explain everything all over again.

Which Kiro feature solves this, and where does it live?

Show model answer

Steering files: markdown files in .kiro/steering/ inside your project. They are loaded into sessions automatically, so the conventions persist without being re-typed.

Steering exists exactly for knowledge that should outlive a single chat: chat context dies with the session, steering does not.
Question 8 · short answer

You want one steering rule applied in every project on your machine, and another applied only inside the current project. Where does each file go?

Show model answer

The machine-wide rule goes in ~/.kiro/steering/; the project-specific one goes in the workspace's .kiro/steering/ folder. If the two conflict, the workspace file wins.

Workspace-over-global precedence lets a project override your personal defaults wherever the project has stronger opinions.
Question 9 · multiple choice

What does a Kiro hook do?
Question 10 · scenario

After every code change, you type the same request: "run the tests and update the matching test file." Kiro does it well, but you are tired of asking, and sometimes you forget to.

What should you set up so this happens without you asking?

Show model answer

A hook triggered on file save. Hooks run an agent prompt or shell command automatically when the event fires, so the tests get run and updated on every save, including the times you would have forgotten.

The valuable part of a hook is that it does not depend on you remembering; that is the difference between automation and a habit.
Question 11 · multiple choice

What does MCP, the Model Context Protocol, let you do in Kiro?
Question 12 · multiple choice

What is the real difference between Autopilot and Supervised mode in chat?
Question 13 · multiple choice

An Autopilot session has made a mess of your files. Which built-in Kiro feature is designed to roll the session back?

Answer key with explanations (all 13 questions)

What is Kiro?

Answer: B. An agentic AI development tool from AWS that plans and writes code through specs, steering, and hooks

Kiro is AWS's agentic development tool: the agent itself plans and edits code, guided by specs, steering, and hooks. Hosting, linting, and ticket tracking are jobs it might help with, but none of them is what Kiro is.
Kiro ships on three surfaces. Which list is correct?

Answer: A. A desktop IDE, the kiro-cli terminal tool, and Kiro Web in the browser

The three surfaces are the IDE (a standalone fork of Code OSS), the CLI, and Kiro Web at app.kiro.dev. The IDE is its own application, not an extension you add to another editor. That mix-up is easy to make, since it is built on the open-source base of VS Code.
Who builds Kiro, and roughly when did it first appear?

Answer: Kiro is built by AWS (Amazon). It launched in public preview in July 2025 and reached general availability in November 2025.

Knowing the source matters: Kiro sits in the AWS family, which is why AWS Builder ID is a sign-in option and why the Amazon Q Developer CLI was later folded into Kiro CLI.
When do Kiro's own docs say a spec session beats a vibe chat session?

Answer: C. For complex features and risky bugfixes, where a reviewable plan is worth the extra structure

Specs trade speed for structure: requirements, design, and tasks you can review before any code is written. That pays off on complex or risky work; for exploration and quick prototypes, vibe chat is the better fit.
A feature spec produces three markdown files. Which one holds the step-by-step plan Kiro actually executes?

Answer: B. tasks.md

tasks.md is the execution checklist Kiro works through. requirements.md says what to build and design.md says how to build it; steps.md does not exist.
Name the three files a feature spec generates, and say in one phrase what each holds.

Answer: requirements.md holds what to build, written as testable acceptance criteria; design.md holds the technical approach; tasks.md holds the broken-down implementation plan.

The three files mirror the workflow phases (Requirements, then Design, then Tasks), so each phase leaves a reviewable artifact behind.
Which Kiro feature solves this, and where does it live?

Answer: Steering files: markdown files in .kiro/steering/ inside your project. They are loaded into sessions automatically, so the conventions persist without being re-typed.

Steering exists exactly for knowledge that should outlive a single chat: chat context dies with the session, steering does not.
You want one steering rule applied in every project on your machine, and another applied only inside the current project. Where does each file go?

Answer: The machine-wide rule goes in ~/.kiro/steering/; the project-specific one goes in the workspace's .kiro/steering/ folder. If the two conflict, the workspace file wins.

Workspace-over-global precedence lets a project override your personal defaults wherever the project has stronger opinions.
What does a Kiro hook do?

Answer: D. Automatically runs an agent prompt or shell command when an event fires, such as a file being saved

Hooks are event-driven automation: file save, create, or delete, prompt submission, tool invocations, and spec task boundaries can all trigger a prompt or command. Connecting external tools is MCP's job, and persistent instructions belong in steering.
What should you set up so this happens without you asking?

Answer: A hook triggered on file save. Hooks run an agent prompt or shell command automatically when the event fires, so the tests get run and updated on every save, including the times you would have forgotten.

The valuable part of a hook is that it does not depend on you remembering; that is the difference between automation and a habit.
What does MCP, the Model Context Protocol, let you do in Kiro?

Answer: B. Connect external tools and data sources, like databases or third-party APIs, so the agent can use them

MCP is a standard plug for outside capability: a server exposes tools and data, and the agent calls them like built-ins. Model choice, settings sync, and credit limits are all handled elsewhere.
What is the real difference between Autopilot and Supervised mode in chat?

Answer: A. Supervised pauses after each agent turn so you can accept or reject file edits hunk by hunk; Autopilot works end to end on its own

The modes differ in when you review, not in what the agent can do. Autopilot acts autonomously but everything stays viewable and revertible; Supervised inserts an approval gate after every turn.
An Autopilot session has made a mess of your files. Which built-in Kiro feature is designed to roll the session back?

Answer: C. Checkpoints: restore points you can return to

Checkpoints are restore points Kiro keeps so agent work can be rolled back confidently. /context clear only changes what the agent sees next, switching modes affects future turns, and .kiro/ holds configuration; deleting it would not undo edits.

Advanced quiz

Fourteen questions for developers past the basics: EARS notation, steering inclusion modes, MCP precedence, custom agents, skills versus powers, and the CLI's sharper edges. These test decisions, not definitions.

Question 1 · multiple choice

Which of these belongs in requirements.md as a well-formed EARS requirement?
Question 2 · multiple choice

In EARS notation as Kiro uses it, which keyword pattern is reserved for error conditions?
Question 3 · scenario

A bug in production is corrupting order totals. The fix touches payment code, so your team wants the root cause, the fix approach, and the verification steps written down and reviewed before anyone edits a line.

Which kind of spec fits, and how does its file set differ from a feature spec?

Show model answer

A bugfix spec. It keeps the spec structure, but the first artifact is bugfix.md instead of requirements.md (documenting the bug rather than feature requirements), followed by design.md and tasks.md.

Kiro's guidance is to reach for specs precisely on risky bugfixes: when the blast radius is payments, the documented, reviewable plan is the point.
Question 4 · short answer

When Kiro executes a tasks.md plan, what does it mean to run tasks in "waves"?

Show model answer

Tasks that do not depend on each other run concurrently as a wave, rather than strictly one at a time; tasks with dependencies wait for an earlier wave to finish.

This is why well-factored, independent tasks in tasks.md complete faster: they parallelize, while a chain of dependent tasks cannot.
Question 5 · multiple choice

You wrote api-conventions.md as a steering file and want it loaded only when the agent works on files under src/api/. Which inclusion mode do you set in its front matter?
Question 6 · multiple choice

The same MCP server is configured in your project's .kiro/settings/mcp.json and in ~/.kiro/settings/mcp.json with different settings. What does Kiro do?
Question 7 · multiple choice

What does listing a tool in an MCP server's autoApprove array actually change?
Question 8 · multiple choice

In a kiro-cli custom agent's JSON definition, which pair of fields controls tool access?
Question 9 · multiple choice

What makes a Kiro power different from adding the same MCP server to mcp.json yourself?
Question 10 · scenario

You maintain a release-notes procedure: a step-by-step writing process plus a helper script. You want every project on your machine to have it, you want it to kick in by itself when you ask Kiro to "draft the release notes", and you want to be able to invoke it explicitly by name.

Steering, a skill, or a power: which one is this, and why?

Show model answer

A skill. Skills package instructions and optional scripts in a SKILL.md whose name and description frontmatter drive activation: Kiro discovers them at session start, auto-activates them when a request matches the description, and accepts explicit /skill-name invocation. Put it in ~/.kiro/skills/ to make it global. Steering is passive knowledge without scripts, and powers exist to bundle a technology's MCP tooling.
Question 11 · scenario

Autopilot has been refactoring your auth module for twenty minutes. The first few turns were exactly right, but the direction it took three turns ago is wrong, and you do not want to throw away the good early work.

Which feature rescues this session, and how?

Show model answer

Checkpointing. Kiro keeps restore points as the agent works, so you roll back to the checkpoint just before the wrong turn and keep everything before it, instead of reverting the whole session.

Checkpoints are what make Autopilot a reasonable default: autonomy is cheap to grant when any stretch of it can be unwound precisely.
Question 12 · short answer

Name the three ways to resume a previous kiro-cli chat session, and say when each fits.

Show model answer

kiro-cli chat --resume reopens your most recent conversation, --resume-id <id> reopens one specific conversation when you know its id, and --resume-picker shows a list of past sessions to choose from.

Resuming carries the old conversation's context forward, which beats starting fresh and re-explaining everything.
Question 13 · short answer

You need kiro-cli to run unattended in a CI pipeline. What does Kiro provide for authentication when nobody can complete an interactive sign-in?

Show model answer

Headless usage with API-key authentication: the pipeline supplies an API key instead of a browser sign-in, so the CLI can run without a human present.

The interactive identity flows (Google, GitHub, AWS Builder ID, SSO) assume a person at the keyboard; an API key is the credential shape that survives automation.
Question 14 · multiple choice

Your automation scripts still call q chat from the Amazon Q Developer CLI. After the upgrade to Kiro CLI, what happens?

Answer key with explanations (all 14 questions)

Which of these belongs in requirements.md as a well-formed EARS requirement?

Answer: C. WHEN a user submits the signup form with an empty email field, THE SYSTEM SHALL display an inline validation error

EARS pairs a concrete trigger with an observable behavior: WHEN <condition> THE SYSTEM SHALL <response>. The others fail as requirements. "Intuitive" and "appropriately" cannot be tested, and the user-story format states a goal rather than a verifiable system behavior.
In EARS notation as Kiro uses it, which keyword pattern is reserved for error conditions?

Answer: B. IF / THEN

IF/THEN marks error handling, WHEN/THE SYSTEM SHALL covers normal event-driven behavior, and WHILE covers ongoing states. GIVEN/WHEN/THEN comes from BDD testing and is not an EARS pattern.
Which kind of spec fits, and how does its file set differ from a feature spec?

Answer: A bugfix spec. It keeps the spec structure, but the first artifact is bugfix.md instead of requirements.md (documenting the bug rather than feature requirements), followed by design.md and tasks.md.

Kiro's guidance is to reach for specs precisely on risky bugfixes: when the blast radius is payments, the documented, reviewable plan is the point.
When Kiro executes a tasks.md plan, what does it mean to run tasks in "waves"?

Answer: Tasks that do not depend on each other run concurrently as a wave, rather than strictly one at a time; tasks with dependencies wait for an earlier wave to finish.

This is why well-factored, independent tasks in tasks.md complete faster: they parallelize, while a chain of dependent tasks cannot.
You wrote api-conventions.md as a steering file and want it loaded only when the agent works on files under src/api/. Which inclusion mode do you set in its front matter?

Answer: D. fileMatch

fileMatch keys inclusion to file paths, which is exactly what "only when touching these files" means. always (the default) loads it everywhere, manual waits for an explicit #api-conventions reference, and auto matches the request's description rather than file paths.
The same MCP server is configured in your project's .kiro/settings/mcp.json and in ~/.kiro/settings/mcp.json with different settings. What does Kiro do?

Answer: A. Uses the workspace definition, because workspace config takes precedence over user config

Workspace beats user on conflict. That lets a repository pin server settings for everyone who opens it, while your user-level file supplies personal defaults in projects that do not care.
What does listing a tool in an MCP server's autoApprove array actually change?

Answer: B. That tool runs without a per-call approval prompt: calls execute the moment the agent decides to make them

autoApprove (or "*" for everything) removes the human approval step for those calls. That is sensible for trusted read-only tools and dangerous for anything that mutates data, because the agent acts without you in the loop.
In a kiro-cli custom agent's JSON definition, which pair of fields controls tool access?

Answer: D. tools and allowedTools

tools defines what the agent can reach, and allowedTools marks which of those run without stopping for approval. prompt and resources shape context, model picks the engine, and there are no permissions or scopes fields.
What makes a Kiro power different from adding the same MCP server to mcp.json yourself?

Answer: A. A power bundles the server config with a POWER.md that tells the agent what tools it has and when to use them, and it loads on demand when keywords match

The power format is about packaging and context economy: POWER.md steering plus MCP config (and optional steering or hooks) that loads only when relevant keywords come up, costing less standing context than an always-connected server. Stripe, Supabase, and Figma were among the launch powers.
Steering, a skill, or a power: which one is this, and why?

Answer: A skill. Skills package instructions and optional scripts in a SKILL.md whose name and description frontmatter drive activation: Kiro discovers them at session start, auto-activates them when a request matches the description, and accepts explicit /skill-name invocation. Put it in ~/.kiro/skills/ to make it global. Steering is passive knowledge without scripts, and powers exist to bundle a technology's MCP tooling.
Which feature rescues this session, and how?

Answer: Checkpointing. Kiro keeps restore points as the agent works, so you roll back to the checkpoint just before the wrong turn and keep everything before it, instead of reverting the whole session.

Checkpoints are what make Autopilot a reasonable default: autonomy is cheap to grant when any stretch of it can be unwound precisely.
Name the three ways to resume a previous kiro-cli chat session, and say when each fits.

Answer: kiro-cli chat --resume reopens your most recent conversation, --resume-id <id> reopens one specific conversation when you know its id, and --resume-picker shows a list of past sessions to choose from.

Resuming carries the old conversation's context forward, which beats starting fresh and re-explaining everything.
You need kiro-cli to run unattended in a CI pipeline. What does Kiro provide for authentication when nobody can complete an interactive sign-in?

Answer: Headless usage with API-key authentication: the pipeline supplies an API key instead of a browser sign-in, so the CLI can run without a human present.

The interactive identity flows (Google, GitHub, AWS Builder ID, SSO) assume a person at the keyboard; an API key is the credential shape that survives automation.
Your automation scripts still call q chat from the Amazon Q Developer CLI. After the upgrade to Kiro CLI, what happens?

Answer: C. They keep working: the q and q chat entry points are preserved, and config migrates from ~/.aws/amazonq to ~/.kiro automatically

The migration was designed not to break existing workflows: the q entry points survive the rebrand and configuration moves to ~/.kiro on its own. There is no separate legacy environment to fall back to.