Week 11 · Quiz

Week 11 Quiz — Agents & Claude Cowork I

Using Artificial Intelligence · AI 101 Fall 2026 · Prof. Quinn Fictional sample

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Covers: what an agent is · agent vs. chatbot · Claude Cowork (desktop platform, requirements, tasks) · projects (instructions, connected folders, context, memory) · permission modes · safe-use principles
Format: 10 auto-graded items (multiple-choice, matching, multiple-answer, true/false) · 10 points (1 each) · allowed attempts: 1 · No AI on this quiz.

This is the human-readable quiz with its vetted answer key, feedback, blueprint, and Canvas placement block. The import-ready Classic QTI 1.2 is in F-quiz-week-11-qti.xml (generated by the validated build script — parses as 10 items). Reminder: AI is not permitted on Quiz 11 — this check confirms that you understand the Week 11 concepts.

Product-accuracy gate: every claim in this quiz about Claude Cowork was verified against the official Anthropic documentation (support.claude.com) as of 2026-06-29. No feature, plan tier, or capability is asserted without a documentation source.

Questions, key, and feedback

Q1 (MC). An AI agent is best described as —
- A. a chatbot that answers questions in a single reply
- B. software that takes multi-step actions on your behalf to accomplish a goal ✅
- C. a search engine that uses AI to rank results
- D. a database of AI-generated answers stored for reuse

Feedback: An agent takes multi-step actions — it plans, executes, and delivers. A chatbot replies once per prompt and you do the rest. Search engines (C) and answer databases (D) are different categories of software entirely.

Q2 (MC). Which statement best distinguishes an agent from a standard chatbot?
- A. Agents are always connected to the internet; chatbots never are
- B. A chatbot replies within a single turn; an agent plans and executes a sequence of actions across multiple steps ✅
- C. Agents can only work with text; chatbots handle images too
- D. A chatbot and an agent are two names for the exact same thing

Feedback: The defining difference is action vs. reply. Option D is the classic misconception — "an agent is just a chatbot" — and it's wrong. A and C describe differences that don't define the agent/chatbot distinction.

Q3 (Matching). Match each Claude Cowork term to its correct meaning.

Term	Correct meaning
Project	A persistent workspace with its own files, instructions, and memory
Connected folder	A local folder on your computer that Claude can read and write
Task	A prompt you give Claude in Cowork that it executes as multi-step work
Global instructions	Standing preferences (tone, format) that apply to every Cowork session

Feedback: These four terms are the Week 11 core vocabulary. "Project" is not a single conversation — it's a persistent workspace. "Connected folder" is your actual local filesystem folder, not an upload. "Task" is the unit of agentic work. "Global instructions" are cross-session, while project instructions are scoped to one project.

Q4 (MC). In Claude Cowork, a "project" is —
- A. exactly the same as a single chat conversation
- B. a persistent, self-contained workspace with its own files, instructions, and memory ✅
- C. a public repository where anyone can see your work
- D. a scheduled task that runs every morning

Feedback: A project is not a chat (A), not public (C), not a scheduled task (D). It's a persistent workspace — meaning it persists across sessions, stores its own files and instructions, and carries memory from one task to the next. Source: Organize your tasks with projects in Claude Cowork.

Q5 (MC). When you connect a local folder in Claude Cowork, it means —
- A. the folder is uploaded to Anthropic's servers permanently
- B. Claude can read files from and write files to that folder on your computer ✅
- C. the folder is shared with all your classmates automatically
- D. Claude cannot make any changes — it can only view the files

Feedback: A connected folder gives Claude read and write access to that folder on your local computer (B). It is not uploaded to any server (A), not shared (C), and not read-only (D). This is why the safety guidance says to connect only the folder Claude needs — no more.

Q6 (Multiple answer — select all that apply). Which of the following are TRUE about Claude Cowork, per the official documentation? Select all that apply.
- A. Cowork requires the Claude Desktop app (macOS or Windows) — it is not available in the browser or on mobile alone ✅
- B. Cowork is available to paid Claude plans (Pro, Max, Team, Enterprise) ✅
- C. Cowork can run tasks even when your computer is asleep and the app is closed
- D. Memory in Cowork is scoped to a project — what Claude learns in one project does not carry to others ✅

Feedback: True: A (Desktop app required), B (paid plans only), D (project-scoped memory). C is false — the official docs state: "Scheduled tasks only run while your computer is awake and the Claude Desktop app is open." This applies to active tasks too: if you close the app or the computer sleeps, the task stops.

Q7 (MC). A student asks Claude Cowork to "summarize my notes" but Claude writes a summary to the wrong location and misses half the files. Which fix is MOST likely to improve the next run?
- A. Switch to a regular chatbot, which is more reliable than an agent for this task
- B. Add project instructions that specify the correct input folder, the output filename and location, and exactly which files to include ✅
- C. Type the request in all capital letters so the agent pays more attention
- D. Ask the agent to summarize faster next time

Feedback: The problem is an under-specified task, not a wrong tool (A), wrong typography (C), or speed issue (D). Agents need precise instructions — file paths, output locations, scope of work. That precision belongs in the project instructions, which persist across all tasks in the project.

Q8 (True/False). An AI agent is just a chatbot with a fancier name — both simply respond to one message at a time and do not take actions outside the chat window.
- True
- False ✅

Feedback: False. This is the week's core misconception. Agents take multi-step actions on your behalf — they can read files, write files, coordinate sub-tasks, and execute work in the background. A chatbot replies to one message; you do the rest. The difference is fundamental.

Q9 (MC). Which statement about memory in Claude Cowork projects is accurate?
- A. Claude's memory is fully global — everything it learns in one project is automatically available in all other projects
- B. Cowork has no memory at all — each task starts completely fresh with no context from prior tasks
- C. Memory is scoped to the project: Claude remembers context from prior tasks within that project and applies it to future tasks in the same project ✅
- D. Memory is only available on Enterprise plans, not on Pro or Max

Feedback: Memory is project-scoped (C) — verified in the official docs: "Memory is scoped to the project, so what Claude learns in one project doesn't carry over to others." A and B are both wrong: memory is real, but bounded. D is wrong: the docs do not restrict memory to Enterprise.

Q10 (MC). Which behavior reflects the correct safety posture when using Claude Cowork on tasks that involve your real files?
- A. Use "Act without asking" mode for every task so Claude can work faster
- B. Grant Claude access to every folder on your computer to ensure it can find what it needs
- C. Use "Ask before acting" mode when working with sensitive or unfamiliar files, and review Claude's plan before letting it proceed ✅
- D. Trust every action Claude takes because it requires your permission before deleting any files anyway

Feedback: "Ask before acting" (C) is the recommended mode for sensitive or unfamiliar files — from the official Use Claude Cowork safely doc. A ("always act without asking") removes oversight. B ("grant access to everything") violates least-privilege. D misreads the docs: Claude does require permission before deleting files, but it can overwrite files without a deletion prompt — so verifying output before acting on it remains essential.

Answer key (quick reference)

Q	Answer
1	B — agent = multi-step actions on your behalf
2	B — chatbot replies once; agent plans and executes a sequence
3	Project → persistent workspace / Connected folder → local folder Claude reads+writes / Task → multi-step prompt execution / Global instructions → cross-session preferences
4	B — persistent workspace with files, instructions, and memory
5	B — Claude reads from and writes to that folder on your computer
6	A, B, D (C is false — tasks stop when the app closes or computer sleeps)
7	B — add project instructions specifying path, output, and scope
8	False — agents take multi-step actions; not just replies
9	C — memory is project-scoped
10	C — "Ask before acting" for sensitive/unfamiliar files

Blueprint & item-bank note

#	Type	Concept	Objective
1	MC	What an agent is	5
2	MC	Agent vs. chatbot	5
3	Matching	Cowork term → meaning (core vocabulary)	5
4	MC	What a Cowork project is	5
5	MC	Connected folder = local read/write	5
6	Multiple answer	Cowork facts (requirements, memory)	5
7	MC	Scenario — prompting fix (better project instruction)	5
8	True/False	Agent ≠ chatbot (misconception)	5
9	MC	Memory is project-scoped	5
10	MC	Safe-use posture (Ask before acting)	5

All 10 items are tagged course=AI101 · week=11 · objective=5 and deposited in the item bank. Distractors target the week's classic misconceptions: "agent = chatbot," "project = a chat," "Cowork runs in the browser," "memory is global," "Act without asking is always fine."

Quality gate (self-checked)

Structure: 10 items, 1 pt each; types = 7 MC + 1 matching + 1 multiple-answer + 1 true/false. ✅
Single-answer integrity: every MC and the true/false have exactly one correct option. ✅
Matching: pairs one-to-one (Project / Connected folder / Task / Global instructions). ✅
Multiple-answer integrity: keys A, B, D (C must be left unselected). ✅
Product-accuracy gate: PASS. Every Cowork feature and constraint cited in this quiz was verified against official Anthropic documentation (support.claude.com) before shipping. Verified docs: Get started with Claude Cowork · Organize your tasks with projects in Claude Cowork · Use Claude Cowork safely. No feature, plan tier, setting, or constraint is invented or asserted without a documentation source.
QTI parse confirmation: F-quiz-week-11-qti.xml parses as imsqti_xmlv1p2 with 10 items; each single-answer respcondition sets SCORE = 100 on exactly one option. ✅

Canvas placement block

canvas_object    = Quizzes::Quiz
title            = "Week 11 Quiz — Agents & Claude Cowork I"
assignment_group = "Quizzes"
points_possible  = 10
grading_type     = points
available_from_offset_days = 0
due_offset_days  = 6
published        = true
allowed_attempts = 1
shuffle_answers  = true
ai_permitted     = false
provenance       = "~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com"

This is the human-readable quiz with its vetted answer key and rationale. The import-ready Classic-QTI version (F-quiz-week-11-qti.xml) ships inside the course's .imscc package — it lands in the Canvas gradebook on import.

~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com