awesome-codex-subagents

Why we included this project

Anyone who works with OpenAI's Codex and keeps wishing it had a sharper specialist for a particular task will find this collection useful. It gathers 171+ ready-made subagent definitions, each a small TOML file that spells out a role, its instructions, a model choice, and a sandbox mode, grouped into categories like core development, language specialists, and quality and security. Instead of writing a fresh prompt every time, you copy the agent file you want into your global or project-level Codex agents directory and call it by name. The defaults are sensible: reviewers and auditors run in read-only sandboxes, builders get workspace-write access, and model routing is tuned per task. It's a good base for teams that want their Codex sessions to behave consistently across different kinds of engineering work.