The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of the codex command-line utility. It provides specific subcommands and flags for various tasks such as analysis (codex exec), code review (codex review), and session management (codex resume, codex fork).
[SAFE]: The instructions prioritize safety by defaulting to a read-only sandbox mode (--sandbox read-only) for general analysis and review tasks.
[SAFE]: High-risk operations, such as modifying workspace files (--sandbox workspace-write) or bypassing security approvals (--dangerously-bypass-approvals-and-sandbox), are strictly gated behind requirements for explicit user consent.
[SAFE]: The skill includes explicit guidance for the AI to treat Codex outputs as non-authoritative and to verify information against primary documentation, mitigating risks from potential tool errors or hallucinations.

codex