Review a specific commit:

codex review --commit <sha>

Review with custom instructions:

codex review "Focus on security vulnerabilities"

Run a task non-interactively:

codex exec "<prompt>"

With full auto mode (sandboxed automatic execution):

codex exec --full-auto "<prompt>"

With specific model:

codex exec -m o3 "<prompt>"

Available Commands

Command	Description
`codex review`	Run code review non-interactively
`codex exec`	Run tasks non-interactively
`codex`	Interactive mode
`codex apply`	Apply latest diff from Codex agent
`codex resume`	Resume previous session

Review Command Options

Option	Description
`--uncommitted`	Review staged, unstaged, and untracked changes
`--base <branch>`	Review changes against base branch
`--commit <sha>`	Review changes from a specific commit
`--title <title>`	Optional commit title for review summary
`-c model=<model>`	Use specific model

Option	Description
`-m, --model <model>`	Model to use (e.g., o3, gpt-4)
`-s, --sandbox <mode>`	Sandbox: read-only, workspace-write, danger-full-access
`--full-auto`	Low-friction sandboxed automatic execution
`-C, --cd <dir>`	Working directory
`-i, --image <file>`	Attach image(s) to prompt
`-o, --output-last-message <file>`	Write final response to file
`--json`	Output events as JSONL

Review current changes before committing:

codex review --uncommitted

Review all changes in a feature branch:

codex review --base main --title "Feature: Add user authentication"

Review with specific focus:

codex review --uncommitted "Focus on:
1. Security vulnerabilities
2. Error handling
3. Performance issues"

Run a coding task:

codex exec "Add input validation to all API endpoints in src/api/"

Get explanation of complex code:

codex exec "Explain the algorithm in src/utils/ranking.R"

Investigate test failures:

codex exec "Investigate why tests in tests/unit/ are failing"

Mode	Permissions
`read-only`	Can read files, no writes
`workspace-write`	Can write to workspace (default with --full-auto)
`danger-full-access`	Full system access (use with caution)

When using this skill:

Use Codex for initial review:

codex review --uncommitted > review.txt

Have Claude Code analyze and summarize:
- Read the review output
- Identify critical issues
- Suggest fixes

Use --full-auto for routine tasks - Speeds up execution with safe defaults
Specify --base for PR reviews - Ensures complete diff is reviewed
Add context to review prompts - Better reviews with specific focus areas
Check sandbox mode - Use appropriate permissions for the task
Save output for complex reviews - Use -o flag for later analysis