Letting Claude Code Run on Its Own (Carefully)

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-builders-agentic-letting-claude-code-loose-r7a8-teen

1. What is Claude Code?

   1. A cloud service that hosts code repositories remotely
   2. A code editor that highlights syntax errors in real-time
   3. A debugging tool that steps through code line by line
   4. An Anthropic terminal agent that can edit files, run tests, and fix errors while unattended

2. What does it mean to 'fence in' an AI agent like Claude Code?

   1. Training the agent to recognize different programming languages
   2. Setting strict boundaries on which files and commands the agent can access
   3. Completely turning off the agent's network access
   4. Installing additional security software on your computer

3. Why is it important to commit your code to git before running an autonomous agent?

   1. So you can undo any unwanted changes if the agent behaves unexpectedly
   2. To share your code with team members before the agent modifies it
   3. To make the agent run faster by caching the code
   4. To automatically fix any syntax errors in your code

4. What is the purpose of an allowed-commands list when running Claude Code?

   1. To define the exact output format the agent should produce
   2. To specify which team members can watch the agent work
   3. To permit only safe, non-destructive commands while blocking dangerous ones like rm -rf
   4. To list all the programming languages the agent is allowed to use

5. What does the lesson mean by the 'agentic mindset'?

   1. Reading documentation carefully before writing any code
   2. Using an AI as a teammate that can make decisions and take action on your behalf
   3. Writing detailed comments in your code for future reference
   4. Testing your code multiple times before committing

6. What is a recommended practice before letting Claude Code run unattended on a project?

   1. Give the agent full access to your entire computer system
   2. Let the agent decide what tasks to work on without any direction
   3. Skip using version control since the agent will handle everything
   4. Run the agent in a fresh or throwaway folder and give it one bounded task

7. If Claude Code makes unwanted changes to your project, what git command can restore your code to its previous state?

   1. git reset --hard to discard all uncommitted changes
   2. git commit -m to save the current changes
   3. git merge to combine different versions
   4. git branch to create a new development line

8. Why should you run an autonomous agent in only one folder rather than your entire system?

   1. To make the agent run faster by reducing its search scope
   2. To contain any damage the agent might cause to just that folder
   3. To ensure the agent can only write code in one programming language
   4. To prevent the agent from accessing the internet

9. What is a 'guardrail' in the context of autonomous AI agents?

   1. A documentation file that explains how to use the agent
   2. A visual interface that shows the agent's progress
   3. A type of automated test that checks code quality
   4. A safety restriction that prevents the agent from performing certain actions

10. What is the danger of giving an autonomous agent like Claude Code full access to your entire project?

    1. The agent will run much slower than usual
    2. The agent will refuse to work on any task
    3. The agent will automatically fix all bugs without asking
    4. The agent might accidentally delete important files or make unintended changes

11. What does it mean for Claude Code to work 'unattended'?

    1. The agent requires your approval for every single action it takes
    2. The agent automatically backs up your files every minute
    3. The agent continues working on tasks without you needing to monitor it constantly
    4. The agent runs without any internet connection

12. What is the benefit of setting up 'permissions' for Claude Code?

    1. It allows the agent to access premium features of your IDE
    2. It automatically generates documentation for your code
    3. It restricts what the agent can do, preventing accidental harm to your project
    4. It makes the agent run faster by prioritizing certain tasks

13. Why is the phrase 'letting Claude Code loose' potentially concerning from a safety perspective?

    1. It indicates the agent will refuse to communicate with you
    2. It suggests the agent will work too slowly to be useful
    3. It means the agent will only work on small, trivial tasks
    4. It implies running the agent without proper restrictions or boundaries in place

14. What should you do if you want to experiment safely with Claude Code for the first time?

    1. Install it on a throwaway project, commit your code, and give it a simple bounded task
    2. Run it on your main production project to see real results
    3. Skip using git so you can see exactly what changes the agent makes
    4. Give it a complex task immediately to see what it can do

15. What is the relationship between 'autonomy' and 'guardrails' when working with AI agents?

    1. Autonomy eliminates the need for any safety restrictions
    2. Guardrails enable safer autonomy by defining what the agent can and cannot do
    3. Guardrails make agents completely dependent on human input
    4. Guardrails slow down agents to the point of being useless