codex-exec

Codex CLI

Use this skill when you need the local codex CLI from a terminal harness. Bias toward non-interactive runs. Use the interactive Codex UI only when the user explicitly wants to stay inside it.

First Check

Confirm the CLI is available before you build a workflow around it:

codex --version

If that fails, stop and report the setup issue instead of retrying the task.

Default Stance

Unless the user asks for something else:

• Use codex exec for one-shot work.
• Use codex review for code-review requests.
• Use codex exec resume --last to continue the most recent saved session.
• Use gpt-5.4 as the default model.
• Use medium reasoning for ordinary work, high for harder tasks, and low for tiny checks.
• Use --sandbox read-only for codex exec analysis runs, and use codex review for review tasks.
• Expand to workspace-write or --full-auto only when Codex should edit files.
• Use --json or --output-schema only when another tool will parse the result.

Do not hide stderr by default, and do not add --skip-git-repo-check unless you are intentionally running outside a Git repo.

For the current flag surface and example command matrix, see ./references/codex-cli.md.

Pick The Run Shape

One-shot run

Default choice for most delegated work:

codex exec \
  --model gpt-5.4 \
  --sandbox read-only \
  -c model_reasoning_effort="medium" \
  "Summarize the uncommitted changes"

Code review

Use the dedicated review subcommand instead of hand-rolling a review prompt through exec:

codex review --uncommitted \
  "Focus on bugs, regressions, and missing tests. Findings first."

Resume the latest session

Use stdin when the follow-up prompt is large or multi-line:

printf '%s\n' "Continue and focus on the failing tests only" | \
  codex exec resume --last -

When resuming, keep prior settings unless the user explicitly wants to change model, sandbox, or autonomy level.

Codex-Specific Gotchas

• codex review is the default review path when the task is code review.
• codex exec reads instructions from stdin when the prompt argument is omitted or set to -.
• codex exec resume --last is the non-interactive continuation path; do not replace it with the top-level interactive codex resume command.
• --full-auto is only a convenience alias for editable autonomous runs. It is not appropriate for read-only analysis.
• --skip-git-repo-check is a situational escape hatch, not a default.
• --ephemeral is useful for disposable runs when session persistence would add noise.

Structured Output

Use structured output only when another tool needs to consume the result:

codex exec \
  --model gpt-5.4 \
  --sandbox read-only \
  --output-schema schema.json \
  "Review the current diff"

Troubleshooting

• If codex --version fails, report a CLI issue.
• If codex exec or codex review exits non-zero, treat the run as failed.
• If a non-interactive run needs too much context, switch from argv text to stdin.
• If the task is review-oriented, use codex review before inventing a custom exec prompt.
• If a run needs edits, change the sandbox and autonomy deliberately instead of piling on flags by habit.
• If output includes warnings or partial results, summarize what Codex completed and what remains uncertain.

Update Check

On first use in a session, silently check for a newer version:

1. Fetch https://raw.githubusercontent.com/abpai/skills/main/versions.json.
2. Compare the version for codex-exec against this file's metadata.version.
3. If the remote version is newer, pause before the main task and ask:

   codex-exec update available (local {X.Y} → remote {A.B}). Would you like me to update it for you first? I can run npx skills update codex-exec for you.

4. If the user says yes, run the update before continuing.
5. If the user says no, continue with the current local version.
6. If the fetch fails or web access is unavailable, skip silently.