Build fluency in the terminology behind OpenAI's terminal-based coding agent.
0 / 5 completed
1 / 5
At standup, a dev wants to run an OpenAI coding agent locally from the terminal against their own repo. Which tool fits?
Codex CLI is OpenAI's terminal-based coding agent that reads, edits, and runs commands in a local repository on the developer's behalf. It is designed for local, sandboxed autonomy rather than a hosted chat widget. This distinguishes it from browser-based coding assistants.
2 / 5
During a design review, the team wants the agent to propose file edits but never run shell commands without asking first. Which Codex CLI concept controls this?
Codex CLI exposes an approval mode (alongside a sandbox policy) that governs whether file edits and shell commands run automatically, require confirmation, or are blocked entirely. Stricter modes reduce risk when the agent operates on unfamiliar code. Choosing the right mode balances autonomy against safety.
3 / 5
In a code review, a dev notices Codex CLI ran commands in an isolated environment with restricted filesystem/network access. What is this called?
Codex CLI executes commands inside a sandbox that limits filesystem and network access, containing the blast radius of an unintended or incorrect action. This is critical because the agent can execute arbitrary shell commands. Sandboxing is a core safety mechanism for autonomous coding agents.
4 / 5
An incident report shows the agent's response referenced files that no longer existed in the repo. What likely happened?
If the underlying repository changes mid-session (e.g., another process modifies files), the agent's context can go stale and reference outdated file state. Restarting the session or re-reading the current state resolves the mismatch. This is a common pitfall when running agents alongside other concurrent edits.
5 / 5
During a PR review, a teammate wants the agent to keep working across multiple turns without re-explaining the task each time. What supports this in Codex CLI?
Codex CLI maintains a persistent session that carries prior turns, edits, and command output forward, so the agent can iterate on a task without the developer re-supplying context. This statefulness is what enables multi-step autonomous work. It mirrors how a human developer keeps context while working through a task.