The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to invoke a local script located at ${CLAUDE_PLUGIN_ROOT}/scripts/codex using the shell to execute tasks in the background. It employs secure practices by passing prompts via stdin redirection rather than inline arguments to avoid shell expansion issues.
[EXTERNAL_DOWNLOADS]: The instructions describe how the underlying agents-bridge component may resolve packages via NPM. It encourages the use of the AGENTS_BRIDGE_CODEX_VERSION environment variable to pin specific versions, which is a best practice for maintaining environment integrity during multi-step operations.
[DATA_EXPOSURE]: The skill utilizes the standard /tmp/ directory to store temporary prompt files and execution logs (JSONL). These files are used to manage communication between Claude and the Codex executor and do not involve sensitive system or credential files.

codex-orchestrate