The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the codex binary via subprocess.run in scripts/codex_context_window.py to inspect the model catalog and validate configuration changes. The arguments are handled safely as a list of strings, and the commands are limited to diagnostic and configuration validation tasks (codex doctor, codex debug models).
[SAFE]: The skill demonstrates high-quality security engineering for file system operations:
Atomic Writes: Uses temporary files and os.replace to ensure that configuration updates are atomic and to prevent file corruption during writes.
Automatic Backups: Creates content-addressed backups in a designated backups directory before any modifications are made to the config.toml file.
Strict Validation: Employs the codex tool's own --strict-config doctor utility to verify the integrity of the generated configuration, performing an automatic rollback to the prior state if any errors are detected.
Least Privilege: Ensures that any new configuration files created are set with private permissions (0o600) to prevent unauthorized access by other local users.

codex-1m-context-window-setup