Sandboxing dangerous operations in Agentic AI - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to safely execute a command in a sandboxed environment.

Agentic AI

sandbox.execute([1])

Drag options to blanks, or click blank then click option'

A'ls -la'

B'shutdown now'

C'rm -rf /'

D'format C:'

Attempts:

3 left

💡 Hint

Common Mistakes

Choosing commands that can harm the system like 'rm -rf /'.

Trying to run shutdown or format commands.

The command 'ls -la' lists files and is safe to run in a sandbox. Other commands are dangerous and should not be executed.

2fill in blank

medium

Complete the code to restrict the sandbox to only allow read operations.

Agentic AI

sandbox.set_permissions([1])

Drag options to blanks, or click blank then click option'

A'read-write'

B'execute-only'

C'read-only'

D'no-access'

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'read-write' which allows modifying files.

Using 'execute-only' which allows running code.

Setting permissions to 'read-only' ensures the sandbox can only read files, preventing dangerous write or execute operations.

3fill in blank

hard

Fix the error in the sandbox initialization to prevent dangerous code execution.

Agentic AI

sandbox = Sandbox([1]=False)

Drag options to blanks, or click blank then click option'

Aallow_file_write

Benable_debug

Cenable_logging

Dallow_network

Attempts:

3 left

💡 Hint

Common Mistakes

Disabling network instead of file write permissions.

Disabling logging or debug which do not affect safety.

Setting 'allow_file_write=False' disables file writing in the sandbox, preventing dangerous modifications.

4fill in blank

hard

Fill both blanks to create a sandbox that limits CPU and memory usage.

Agentic AI

sandbox = Sandbox(cpu_limit=[1], memory_limit=[2])

Drag options to blanks, or click blank then click option'

A50

B1024

C2048

D100

Attempts:

3 left

💡 Hint

Common Mistakes

Setting CPU limit too low or memory limit too high.

Confusing units for CPU and memory limits.

Setting cpu_limit to 100% and memory_limit to 1024 MB restricts resource usage to safe levels.

5fill in blank

hard

Fill all three blanks to safely execute user code with timeout and restricted imports.

Agentic AI

sandbox = Sandbox(timeout=[1], allowed_imports=[2], safe_mode=[3])
result = sandbox.run(user_code)

Drag options to blanks, or click blank then click option'

A5

B['math', 'random']

CTrue

DFalse

Attempts:

3 left

💡 Hint

Common Mistakes

Allowing all imports which can be dangerous.

Disabling safe_mode which reduces protection.

Setting no timeout leading to infinite loops.

Setting a 5-second timeout, allowing only safe imports, and enabling safe_mode ensures secure execution.