All classes Claude from Zero to Expert Syllabus · Episode 59 of 100

Claude from Zero to Expert Class #59: When to Allow an Action and When to Refuse

0 views

A guide to permission decisions for Claude in Chrome: green, yellow, and red lights for agent actions, four questions when a permission box appears, three real scenarios (allowing a draft save, refusing a download and an OTP request, taking over a login), and safe habits such as limiting automatic permissions. Episode 59 of 100, Session 12: Claude in the Browser.

Software: Claude
Level: Intermediate
Duration: ± 8 min
Series total: 15 h 1 min · 100 videos
Published: October 28, 2026

Episode 59 of 100 · view syllabus

Subscribe ▶ Watch the video lesson0 plays

Checking access...

Want to go deeper?

Join one of our courses or ask us directly on WhatsApp.

View the full 100-episode syllabus View Claude AI Courses Ask on WhatsApp

Recommended classes for you

Picked from the software, topic, and level of “Claude from Zero to Expert Class #59: When to Allow an Action and When to Refuse”.

50 episodesFree preview

Agentic AI · Beginner–Intermediate

Agentic AI Without Coding

50 video lessons with PDF guides, 14 h 29 min in total. Starts with: Chatbots, Automation, and AI Agents: What's the Difference?.

14 h 29 min 66 0

View class

6 episodesFree preview

Python · Intermediate–Advanced

Brain Tumor Segmentation with U-Net

6 video lessons with PDF guides, 1 h 10 min in total. Starts with: Brain MRI Dataset & Tumor Mask: Loading, Visualization & Patient-wise Splitting.

1 h 10 min 28 0

View class

Deep Learning CNN-LSTM Pneumonia Detection

6 episodesFree preview

Python · Intermediate–Advanced

Deep Learning CNN-LSTM Pneumonia Detection

6 video lessons with PDF guides, 57 min in total. Starts with: Chest X-ray Dataset: Loading, Exploration & Data Splitting.

57 min 22 0

View class

ANFIS Metaheuristics Time Series Forecasting

6 episodesFree preview

Python · Intermediate–Advanced

ANFIS Metaheuristics Time Series Forecasting

6 video lessons with PDF guides, 59 min in total. Starts with: Non-Linear Time Series: Dataset, Stationarity, and Patterns.

59 min 13 0

View class

Need more than a video?

Other ways we can help with Claude.

Learn with a mentor

Scheduled AI & Machine Learning courses with live guidance and a certificate.

Course

Midjourney Course: Create Stunning AI Images from Text

★ 4.9

Course

DALL-E Course: Creating Professional AI Images from Text

★ 4.9

Course

Stable Diffusion Course: AI Art from Beginner to Professional

★ 4.9

See all in AI & Machine Learning →

Have it built for you

No time to learn Claude? Our team can do the project for you.

Service

AI Image Creation with Stable Diffusion — Unlimited Custom Visual Production

★ 4.9

Service

AI Image & Video Creation with Leonardo AI — All-in-One Visuals for Business

★ 4.9

Service

AI Image Creation with Midjourney — Cinematic AI Art for Business

★ 4.9

See all in AI & Machine Learning →

Get notified about new Claude classes

At most one email a week. Unfollow anytime with one click.

Transcript

The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.

Show transcript

Introduction

Hi, and welcome back to Claude from Zero to Expert. In the last three episodes, the agent stopped several times and waited for our decision. Today we learn to decide quickly and safely. We'll build traffic lights for agent actions, use four questions when a permission box appears, and practice with three real scenarios: allowing, refusing, and taking over.

Step 1: Traffic lights for agent actions

When the agent asks whether it may do something, your answer isn't a formality. The agent acts on your behalf, using the accounts that are signed in to the browser. That's why we need a quick, consistent way to decide, so we don't press allow out of boredom, or refuse everything out of fear. Use traffic lights. A green light is for reading, searching, moving between pages, and filling in search boxes. Nothing changes in the real world, so let the agent run. A yellow light is for actions that change something but can still be undone: saving a draft, filling in a form, adding an item to a cart. Allow them after you check the content. A red light is for logins, payments, personal data, sending, deleting, and downloading files. Also any instruction that comes from a web page rather than from you. Refuse, or take over and do it yourself. One simple question separates yellow from red: if it goes wrong, can it be easily undone? If not, it's red.

Step 2: Four questions when a permission box appears

Every time a permission box appears, answer four questions. First, what exactly is the action? Saving, sending, or paying? Read the sentence in the permission box all the way through. Second, on which site? Look at the address in the top bar. A site with a look-alike name could be fake. Third, what data will leave you? Your business name is fine. A phone number, home address, or customer data deserves a second thought. Fourth, if it goes wrong, can it be undone? This is the question that decides the color of the light. If any answer is unclear, refuse for now or ask the agent to explain. Refusing costs nothing. The agent can continue once you understand. Here's a small example of using the four questions. Laras's agent is ready to submit a contact us form on a roastery's site, and asks for permission. Laras doesn't answer right away. She asks first: what data will be sent? The agent explains: the business name, a message about arabica stock, and a WhatsApp number. It turns out the number is Laras's personal one. Laras asks for it to be replaced with the business number, and only then allows it. One simple question keeps her personal number off a marketing list.

Step 3: Yellow light scenario: allow after checking

The first scenario is from Nadia, the social media freelancer. She asks the agent to prepare a weekend post for Sunset Coffee in the scheduling tool she usually uses. The site is fictional. Nadia already signed in herself earlier. The agent chooses the Sunset Coffee client and the Friday slot at seven in the evening. Then it writes the cold brew caption from the notes Nadia gave it. The agent stops and asks permission to save it as a draft. Let's use the four questions. The action is saving a draft. The site is Nadia's subscribed scheduling tool. No personal data leaves. And a draft can be changed or deleted. This is a yellow light. Nadia checks the date and caption on screen. Everything is right, so she allows it, for this action only. The draft is saved. Look at the last line: the publish button wasn't touched. The post only goes live after Laras, the client, approves, and Nadia presses it herself. A yellow light doesn't mean suspicion. It means check first, then allow. Nadia only needed ten seconds to make sure the date and caption were right.

Step 4: Red light scenario: refuse or take over

The second scenario is from Mr. Hendra. He asks the agent to find fraction practice questions for grade five. The agent opens a free materials site full of ads. This site is fictional, but the pattern is very common. The site asks visitors to download an installer app to open the questions. The agent doesn't download it right away. It asks first. Downloading and installing a file from an unknown site is a red light. Mr. Hendra refuses. Practice questions aren't worth the risk of malware on the computer. Then there's a teacher prize offer asking for a phone number and an OTP code. An OTP code is the key to your account. There's no legitimate reason for a worksheet site to ask for it. The agent ignores it and reports it. The agent then looks for another site that shows the questions right on the page. Refusing doesn't make the task fail; it just makes the agent find a safe route. The third scenario is from Dewi. She asks the agent to check the status of October invoices on the Sinar Paper customer portal. The agent opens the portal address Dewi actually gave. The address is checked first, because fake portals often use similar names. The agent reaches the login page and stops. The agent never fills in a password. This is the moment to take over. Dewi types her email and password herself, then hands control back to the agent. Signing in is a red light, but reading the invoices afterward is a green light. The agent reads the invoice list: two are paid, and one is due on November fifth. That page has a pay button. The agent doesn't touch it. It only reads and reports. Payment is Dewi's decision, through the channel she normally uses. Those two scenarios show two ways to handle a red light. Refuse, when the action isn't needed at all. Take over, when the action is needed but must be done by a person. Examples of refusing: unknown file downloads and prizes asking for an OTP. Examples of taking over: signing in, paying, and ticking an agree-to-terms box. After you refuse, the agent finds another way. After you take over, the agent can continue the safe part, such as reading and reporting.

Step 5: Safe permission habits

Finally, five safe habits. First, allow once, not forever. The don't-ask-again option only suits sites you trust completely, and green-light actions. Second, keep a list of trusted sites for your work. An agent working on sites you know is much safer. Third, close tabs for important accounts you don't need, such as internet banking or personal email, before handing a task to the agent. Fourth, stop the agent if its steps go off track. Don't feel awkward; the stop button is there for exactly that. Fifth, review the step history after the task is done. Permission settings in Claude in Chrome can change, so check Anthropic's official page. The screens in this video are simulations. Let's sum up. When a permission box appears, answer four questions: what action, on which site, what data, and can it be undone. From the answers, pick the color of the light: green, yellow, or red. Then decide: let it run, allow after checking, or refuse. For logins and payments, take over and do them yourself, then hand the reading part back to the agent.

Done!

We've built traffic lights for agent actions, used four questions when a permission box appears, and practiced three scenarios: allowing a draft, refusing a download and an OTP request, and taking over a login. In the next episode we close the Claude in the Browser session with full practice: a supplier price research agent for Sunset Coffee, from the list of sites to a recommendation. See you there!

Practice exercises

Try these on your own after watching to lock in what you learned.

Sort ten actions from your work into green, yellow, and red lights
When trying the agent, answer the four questions every time a permission box appears
Write a list of trusted sites the agent may visit for your work

Ask the class material

Type a question or keyword; we find the part of this episode that explains it and jump the video there.

My notes

Notes are saved with the current video time and are only visible to you.

Questions & answers

Stuck on a step? Ask here; the instructor and other learners can answer.

No questions yet.

Your instructor

Cerdas Inovasi Digital Team

Software developers, BIM & 3D engineers, and AI practitioners

53 video classes1,296 episodes

We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.

See our clients →

Memuat

Claude from Zero to Expert Class #59: When to Allow an Action and When to Refuse

0 views

Software

Claude

Level

Intermediate

Duration

± 8 min

Series total

15 h 1 min · 100 videos

Published

October 28, 2026

Episode 59 of 100 · view syllabus

▶ Watch the video lesson0 plays

Transcript

The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.

Show transcript

Introduction

Step 1: Traffic lights for agent actions

Step 2: Four questions when a permission box appears

Step 3: Yellow light scenario: allow after checking

Step 4: Red light scenario: refuse or take over

Step 5: Safe permission habits

Done!

Your instructor

Cerdas Inovasi Digital Team

Software developers, BIM & 3D engineers, and AI practitioners

53 video classes1,296 episodes

See our clients →

Recommended classes for you

Need more than a video?

Transcript

0:03Introduction

0:27Step 1: Traffic lights for agent actions

1:30Step 2: Four questions when a permission box appears

2:48Step 3: Yellow light scenario: allow after checking

3:59Step 4: Red light scenario: refuse or take over

6:22Step 5: Safe permission habits

7:35Done!

Practice exercises

Ask the class material

My notes

Questions & answers

Your instructor

Recommended classes for you

Need more than a video?

Transcript

0:03Introduction

0:27Step 1: Traffic lights for agent actions

1:30Step 2: Four questions when a permission box appears

2:48Step 3: Yellow light scenario: allow after checking

3:59Step 4: Red light scenario: refuse or take over

6:22Step 5: Safe permission habits

7:35Done!

Practice exercises

Ask the class material

My notes

Questions & answers

Your instructor

Introduction

Step 1: Traffic lights for agent actions

Step 2: Four questions when a permission box appears

Step 3: Yellow light scenario: allow after checking

Step 4: Red light scenario: refuse or take over

Step 5: Safe permission habits

Done!

Introduction

Step 1: Traffic lights for agent actions

Step 2: Four questions when a permission box appears

Step 3: Yellow light scenario: allow after checking

Step 4: Red light scenario: refuse or take over

Step 5: Safe permission habits

Done!