What is this?
A curated collection of 170+ specialized subagent definitions for OpenAI Codex, covering development tasks like coding, infrastructure, security, data/AI, and developer experience.
A curated collection of 170+ specialized subagent definitions for OpenAI Codex, covering development tasks like coding, infrastructure, security, data/AI, and developer experience.Is this a tool or just a set of config files?It is a set of .toml configuration files, not an app or script you run; each file defines a subagent that Codex loads from its agent directories.
It is a set of .toml configuration files, not an app or script you run; each file defines a subagent that Codex loads from its agent directories.How do I install and use these subagents?Clone the repo, copy the .toml files you want into ~/.codex/agents/ (global) or .codex/agents/ (project-specific), then restart or refresh your Codex session and explicitly delegate to the subagent in your prompt.
Clone the repo, copy the .toml files you want into ~/.codex/agents/ (global) or .codex/agents/ (project-specific), then restart or refresh your Codex session and explicitly delegate to the subagent in your prompt.Do I need anything besides Codex itself to use these?You just need an existing Codex setup; agent routing configuration can optionally be added under [agents] in .codex/config.toml.
You just need an existing Codex setup; agent routing configuration can optionally be added under [agents] in .codex/config.toml.Will Codex automatically pick the right subagent for me?No, Codex does not auto-spawn custom subagents, so you must explicitly reference the subagent you want in your prompt.
No, Codex does not auto-spawn custom subagents, so you must explicitly reference the subagent you want in your prompt.Does project/global precedence affect which subagent runs?Yes, when a project-level and global subagent share the same name, the project-specific one takes precedence.
Yes, when a project-level and global subagent share the same name, the project-specific one takes precedence.How does model routing work for each subagent?Each .toml file has a model field that routes it to one of three tiers: gpt-5.6-sol for deep reasoning (e.g. security-auditor, architect-reviewer), gpt-5.6-terra for implementation and diagnostics (e.g. docker-expert, test-automator), and gpt-5.6-luna for lightweight search/extraction tasks (e.g. search-specialist, agent-installer).
Each .toml file has a model field that routes it to one of three tiers: gpt-5.6-sol for deep reasoning (e.g. security-auditor, architect-reviewer), gpt-5.6-terra for implementation and diagnostics (e.g. docker-expert, test-automator), and gpt-5.6-luna for lightweight search/extraction tasks (e.g. search-specialist, agent-installer).How are subagents organized?They are grouped into 13 categories such as Core Development, Language Specialists, Infrastructure, Quality & Security, Data & AI, and Developer Experience, each with its own themed set of agents.
They are grouped into 13 categories such as Core Development, Language Specialists, Infrastructure, Quality & Security, Data & AI, and Developer Experience, each with its own themed set of agents.What is sandboxmode and why does it matter?Each subagent sets a sandboxmode field controlling filesystem access: read-only for reviewers and auditors that only analyze, and workspace-write for developers/engineers who need to create or modify files.
Each subagent sets a sandboxmode field controlling filesystem access: read-only for reviewers and auditors that only analyze, and workspace-write for developers/engineers who need to create or modify files.Can I pick which subagents are available per project vs globally?Yes, you can store agents in .codex/agents/ for project-only use or in ~/.codex/agents/ to make them available across all projects, with project agents taking precedence on name conflicts.
Yes, you can store agents in .codex/agents/ for project-only use or in ~/.codex/agents/ to make them available across all projects, with project agents taking precedence on name conflicts.What are similar projects?ruflo, pstack-claude, infisical, evenfire, awesome-agent-skills, static-analysis, Codex-X, coder
ruflo, pstack-claude, infisical, evenfire, awesome-agent-skills, static-analysis, Codex-X, coderWhich topics is it in?toml, codex-skills, terraform, subagents, code-review, chatgpt, kubernetes, testing, agents, documentation, devops, awesome-list
toml, codex-skills, terraform, subagents, code-review, chatgpt, kubernetes, testing, agents, documentation, devops, awesome-list