A · Beginner · Read 10 min · Practice 15 min
Codex is a coding agent that can inspect a project, edit files and run tools in a chosen environment. A model supplies its reasoning; a task is one assignment; a project groups related work and files. Unlike copying a code answer from chat into an editor, an agent can make the change in the workspace you selected.
Desktop · Mobile · CLI · IDE · Cloud
Updated: 2026-09-14
A · Beginner · Read 10 min · Practice 15 min
Check how you signed in before interpreting usage. ChatGPT sign-in uses the Codex allowance associated with that account; API-key authentication in a supported client requires checking API billing separately. This tutorial teaches you where to verify your own allowance rather than freezing a price or message limit that may change.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
A · Beginner · Read 10 min · Practice 15 min
Compare desktop, mobile, CLI, IDE and cloud scenarios, and identify where files and tools run.
Desktop · Mobile · CLI · IDE · Cloud
Updated: 2026-09-20
A · Beginner · Read 12 min · Practice 20 min
Create a practice folder on Windows, macOS or Linux and distinguish current, relative and absolute paths.
Desktop · CLI · IDE
Updated: 2026-09-14
A · Beginner · Read 12 min · Practice 30 min
Use an independent copy of the Small Steps todo website to change its heading and background. Identify the five practice files and compare the diff, core tests and two viewport widths while preserving behavior and data.
Desktop · CLI
Updated: 2026-09-20
A · Beginner · Read 12 min · Practice 20 min
A useful request explains the outcome, available context, boundaries and acceptance checks. 'Improve the page' could mean speed, appearance or accessibility. Convert it into observable behavior before adding more instructions.
Desktop · Mobile · CLI · IDE · Cloud
Updated: 2026-09-14
B · Beginner · Read 12 min · Practice 25 min
Choose an installation platform, sign in, inspect a file, make a one-line change, verify it independently and resume the session. Includes a repeatable two-file exercise and a missing-file test.
CLI
Updated: 2026-09-14
B · Beginner · Read 12 min · Practice 25 min
Install and sign in from PowerShell, check command discovery, read a project and troubleshoot updates.
CLI
Updated: 2026-09-14
B · Beginner · Read 12 min · Practice 25 min
Set up the CLI in macOS Terminal and separate shell, update, directory and permission problems.
CLI
Updated: 2026-09-14
B · Beginner · Read 12 min · Practice 30 min
Distinguish host and WSL tools, paths and accounts, then run the exercise and update from the right environment.
CLI
Updated: 2026-09-14
B · Beginner · Read 15 min · Practice 25 min
The desktop workspace combines tasks, files, terminals and change review. Current official downloads use the name ChatGPT desktop app; choose Codex inside it. Follow the separate installation path for your system. Linux is a preview with a specific support list, not a promise of compatibility with every distribution.
Desktop
Updated: 2026-09-20
B · Everyday work · Read 12 min · Practice 30 min
IDE integration places Codex beside your code for focused explanation and edits. VS Code, Cursor and Windsurf use the Codex extension; Xcode and JetBrains expose their own integration flows. Do not interchange their installation steps.
IDE
Updated: 2026-09-14
C · Beginner · Read 12 min · Practice 20 min
Remote in the ChatGPT mobile app lets you guide Codex work on a connected computer. Official documentation lists iOS and Android clients, with Mac or Windows hosts, subject to rollout. This does not install a full development environment on the phone, and does not establish Linux host support.
Mobile
Updated: 2026-09-14
C · Beginner · Read 12 min · Practice 25 min
Check the available iOS entry point, pairing and authentication, then track a supported task on its host.
Mobile
Updated: 2026-09-14
C · Beginner · Read 12 min · Practice 25 min
Check the Android app, pairing and notifications, select the correct host and verify the task output.
Mobile
Updated: 2026-09-14
C · Beginner · Read 12 min · Practice 25 min
Pair devices, diagnose host availability and verify fresh file reads. Distinguish remote access, device pairing and task execution when stopping work.
Desktop · Mobile
Updated: 2026-09-14
C · Everyday work · Read 15 min · Practice 35 min
Cloud tasks run in a remote environment configured for a repository. They suit longer delegated work, but cannot automatically see uncommitted files on your computer. Creating a pull request is distinct from merging or deploying it.
Cloud
Updated: 2026-09-14
C · Beginner · Read 12 min · Practice 25 min
Track the conversation, host, directory and branch so a handoff continues the intended work.
Desktop · Mobile · CLI · IDE · Cloud
Updated: 2026-09-14
D · Beginner · Read 10 min · Practice 20 min
Markdown uses plain-text markers for headings, lists, links and code, commonly in .md files. README.md explains a project, AGENTS.md supplies agent instructions and SKILL.md defines a reusable skill. An arbitrary MD file is not automatically an instruction source.
Desktop · CLI · IDE
Updated: 2026-09-14
D · Beginner · Read 12 min · Practice 20 min
AGENTS.md supplies project instructions before Codex starts work. Global guidance and files along the project-root-to-working-directory path form an instruction chain. At the same level, AGENTS.override.md takes precedence. Keep rules concrete and verify what was actually loaded.
Desktop · CLI · IDE
Updated: 2026-09-14
D · Everyday work · Read 12 min · Practice 25 min
Trace root and nested instructions, conflicts and overrides, and verify which files were selected.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
D · Everyday work · Read 12 min · Practice 25 min
Separate working rules, project entry points, design requirements and handoff state. Check outdated hypotheses against actual files and tests.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
D · Everyday work · Read 12 min · Practice 25 min
config.toml sets client options, while AGENTS.md describes working instructions. User settings live in Codex home; trusted projects can add .codex/config.toml. CLI overrides, project layers and managed requirements mean a single file does not prove the effective configuration.
Desktop · CLI · IDE
Updated: 2026-09-14
D · Everyday work · Read 12 min · Practice 25 min
Trace ineffective or conflicting settings, change one item at a time and keep a reversible record.
Desktop · CLI · IDE
Updated: 2026-09-14
E · Beginner · Read 12 min · Practice 20 min
Distinguish shell commands and launch flags, interactive slash commands, keyboard shortcuts and natural-language requests. Check the input location: /plan belongs in the Codex interactive client; codex --help runs in the operating-system terminal.
CLI
Updated: 2026-09-14
E · Everyday work · Read 12 min · Practice 20 min
Distinguish new, resumed and forked tasks, locate the right history and verify the directory and remaining work.
Desktop · CLI · IDE
Updated: 2026-09-14
E · Everyday work · Read 12 min · Practice 22 min
Long work needs durable decisions and evidence, not just a long conversation. README explains use, design documents explain choices, handoffs record current progress and AGENTS.md holds ongoing instructions. Do not turn all temporary progress into permanent rules.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
E · Everyday work · Read 12 min · Practice 20 min
Model choice affects available capabilities and usage conditions; reasoning effort affects how much work the agent devotes to a problem. Compare quality on the same small task before increasing effort. Use the options currently exposed by your account rather than assuming a named model is universally available.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
E · Beginner · Read 13 min · Practice 25 min
Plan mode is useful when scope or approach still needs decisions. It produces an implementation plan, not finished code. Define the questions to resolve and check that the plan includes inputs, outputs, constraints and verification.
Desktop · CLI
Updated: 2026-09-14
E · Everyday work · Read 14 min · Practice 20 min
Sandbox rules define technical access to files and networks; approval policy determines when the agent asks before acting. No prompt does not guarantee permission, and read access does not imply write access everywhere.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
F · Everyday work · Read 12 min · Practice 20 min
Keep follow-ups about one outcome in the same task and separate independent outcomes into different tasks. Projects identify files; tasks track assignments. Multiple tasks do not automatically isolate files: use worktrees when isolation is needed.
Desktop
Updated: 2026-09-14
F · Everyday work · Read 14 min · Practice 25 min
Use a read-only workflow to locate entry points, data flow and tests, with file-backed explanations.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
F · Everyday work · Read 14 min · Practice 30 min
Implement Active and Completed filters from start, preserving data and verifying normal and boundary cases.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
F · Everyday work · Read 14 min · Practice 25 min
Debugging starts with a reproducible symptom, narrows causes with evidence and verifies the fix. Give the input, exact step, expected behavior and observed failure instead of only saying the site is broken.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-20
F · Everyday work · Read 15 min · Practice 30 min
Git stores file history, branches organize changes and diffs show what changed. Codex can help, but you must verify that the changes belong to the task. Restoring an old conversation does not restore Git files.
Desktop · CLI · IDE
Updated: 2026-09-14
F · Everyday work · Read 15 min · Practice 30 min
Tests check behavior under specified conditions, review examines the change for defects and a PR presents proposed integration. They complement each other: tests passing does not establish review approval, and an open PR is not a merge.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
G · Advanced · Read 12 min · Practice 25 min
A skill packages a repeatable workflow with instructions and optional resources. The minimum is a directory containing SKILL.md with name and description metadata followed by the procedure and expected output. Installation alone does not prove a task used it.
Desktop · CLI · IDE
Updated: 2026-09-14
G · Advanced · Read 15 min · Practice 30 min
Move repeated logic and large references into useful resources, then verify conditional loading and relative paths.
Desktop · CLI · IDE
Updated: 2026-09-14
G · Advanced · Read 15 min · Practice 30 min
Test format, selection and outcomes separately, using matching and non-matching requests to refine descriptions.
Desktop · CLI · IDE
Updated: 2026-09-14
G · Advanced · Read 15 min · Practice 25 min
A plugin can distribute skills and MCP tools as an installable capability. Installing it, connecting an external account and successfully calling a tool are distinct stages. A listing does not establish access to your data.
Desktop · CLI
Updated: 2026-09-20
G · Advanced · Read 15 min · Practice 20 min
Distinguish installation, account connection, authorization and callable tools before reconnecting or removing a plugin.
Desktop · CLI
Updated: 2026-09-20
G · Advanced · Read 15 min · Practice 20 min
Choose a prompt, rule or handoff template, fill its required fields and use the correct input location.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-20
H · Advanced · Read 15 min · Practice 20 min
MCP connects Codex to tools and data. STDIO servers usually start as local processes; HTTP servers use a URL. Saving configuration only proves it exists: verify startup, authentication and an actual tool response.
Desktop · CLI · IDE
Updated: 2026-09-20
H · Advanced · Read 15 min · Practice 20 min
Inspect process, transport, authentication and tool discovery, preserve errors and verify a minimal operation.
Desktop · CLI · IDE
Updated: 2026-09-20
H · Advanced · Read 15 min · Practice 20 min
A Git worktree gives one repository multiple working directories on different branches. It isolates file edits, but databases, ports and external services may still be shared. File isolation is not full resource isolation.
Desktop · CLI
Updated: 2026-09-20
H · Advanced · Read 15 min · Practice 20 min
Subagents suit independent work that can be reviewed and combined by the main task. They consume additional model and tool usage. Parallel tasks with mutual dependencies or overlapping edits can create more conflict than speed.
Desktop · CLI
Updated: 2026-09-20
H · Advanced · Read 15 min · Practice 20 min
Define independent inputs, edit scopes and evidence, then review subagent results and resolve gaps.
Desktop · CLI
Updated: 2026-09-14
H · Advanced · Read 20 min · Practice 30 min
Integrate independent edits using file and interface contracts, resolve conflicts and verify combined behavior.
Desktop · CLI · IDE
Updated: 2026-09-20
I · Advanced · Read 15 min · Practice 20 min
Visual work needs a reference, the actual result and explicit change criteria. Browser inspection, screenshots and generated concepts provide different evidence. A generated interface is not proof that a workflow works.
Desktop · CLI
Updated: 2026-09-20
I · Advanced · Read 15 min · Practice 20 min
Scheduled work needs a defined execution environment. A desktop task that reads local files requires the computer and app to be available at run time. Cloud work cannot directly read local data you have not supplied.
Desktop
Updated: 2026-09-20
I · Advanced · Read 15 min · Practice 20 min
codex exec runs one non-interactive task for scripts or CI. Normally progress goes to stderr and the final answer to stdout. --json produces JSON Lines events, not a single JSON object.
CLI
Updated: 2026-09-14
I · Advanced · Read 20 min · Practice 25 min
Separate event streams from final results, parse structured output and reject missing or invalid fields.
CLI
Updated: 2026-09-14
I · Advanced · Read 20 min · Practice 30 min
Design a CI job with explicit input, permissions and exit states, preserving artifacts and separating proposals from application.
CLI · Cloud
Updated: 2026-09-14
I · Advanced · Read 15 min · Practice 20 min
Use execution records to detect prior effects, avoid duplicates, repair prerequisites and confirm a schedule has stopped.
Desktop · CLI
Updated: 2026-09-14
J · Advanced · Read 20 min · Practice 60 min
Plan and build the Small Steps task website from brief.md, with adding, completing, deleting, filtering and local persistence. Separate HTML, CSS, data functions, UI events and tests, verify with Node and browser checks, and document restart and recovery steps.
Desktop · CLI · IDE
Updated: 2026-09-14
J · Advanced · Read 20 min · Practice 45 min
Build a CSV cleaner that preserves its input and writes a new output. The exercise trims whitespace, lowercases email, keeps the first valid occurrence of an address and skips blank names or malformed addresses. It reports kept, duplicate and invalid counts. Its format check does not prove a mailbox exists.
Desktop · CLI · IDE
Updated: 2026-09-14
J · Advanced · Read 15 min · Practice 30 min
Establish a baseline, implement one change and deliver traceable regression checks and handoff notes.
Desktop · CLI · IDE · Cloud
Updated: 2026-09-14
J · Advanced · Read 15 min · Practice 20 min
Create fictional test data and restorable copies, and identify information to exclude from prompts, screenshots and repositories.
Desktop · Mobile · CLI · IDE · Cloud
Updated: 2026-09-14
J · Advanced · Read 15 min · Practice 25 min
Record task conditions, model options, time and outcomes to reduce unnecessary retries and excess context.
Desktop · Mobile · CLI · IDE · Cloud
Updated: 2026-09-14
J · Beginner · Read 15 min · Practice 20 min
Start by locating the failing layer. A missing executable suggests installation or PATH; a rejected account suggests authentication; missing files suggest location or permissions; an incorrect result may involve the request or project code. Change one condition at a time.
Desktop · Mobile · CLI
Updated: 2026-09-14