Code Interpreter

What is Code Interpreter?

A code interpreter tool lets an LLM write and execute code—usually in a sandbox—to compute, analyze data, or call libraries. It expands agency and creates code-execution and data-exfiltration risks if poorly sandboxed.

Related Giskard articles

• Function calling security
• Excessive agency case

Secure code interpreters with Giskard — probe tool misuse, sandbox escape attempts, and excessive agency before production. See function-calling security.

Authority: OpenAI Code Interpreter docs.