codex-skill

Non-interactive automation mode for hands-off task execution using Codex with configurable models and sandbox modes.

• Autonomous execution of tasks without approval prompts
• Supports model selection (gpt-5, gpt-5.1, etc.)
• Provides sandbox modes: read-only, workspace-write, danger-full-access
• Enables full automation with optional file editing and external output

• Execute tasks from start to finish without seeking approval
• Stream progress updates during execution
• Provide structured final summary with actionable results
• Supports model selection via -m flag
• Control access via sandbox modes: read-only, workspace-write, danger-full-access
• Full-auto mode (alias for workspace-write) for programming tasks
• Save output to file with -o flag
• Generate JSON output with structured reasoning and metrics

1. 1.Install and verify Codex CLI: `codex --version`
2. 2.Choose model with -m (e.g., gpt-5.1-codex)
3. 3.Set sandbox mode with -s (read-only, workspace-write, danger-full-access)
4. 4.Use --full-auto for recommended programming tasks
5. 5.Save output with -o or enable --json for structured output
6. 6.Verify flag availability before using dangerous options