oil-oil/codex-team-mode / skills/team-mode
team-mode
VerifiedUse when a task may benefit from a bounded subagent for implementation, large-codebase discovery at task start, independent review, requested or pre-commit code simplification, or expert work on complex decisions, modeling, automation, or repeated failures. The main agent chooses whether to delegate and accepts the result. Do not use for simple questions or short work with no meaningful review or delegation.
| File | Kind | Size |
|---|---|---|
| README.md | markdown | 1.8 KB |
| SKILL.md | skill md | 5.2 KB |
| agents/openai.yaml | config | 410 B |
| references/custom-agents.md | markdown | 4.1 KB |
| references/evaluation.md | markdown | 2.3 KB |
| references/explore.md | markdown | 654 B |
| references/interactive-testing.md | markdown | 1.4 KB |
| references/simplify.md | markdown | 2.9 KB |
| scripts/current_model.pyexecutable | script | 3.4 KB |
| scripts/usage_by_model.pyexecutable | script | 36.0 KB |
references/custom-agents.md
View on GitHub# Custom Agent Profiles
Read this reference when installing, changing, or diagnosing Team Mode profiles. Normal dispatch does not need profile inspection.
## Profiles
Copy the templates from the repository's `agents/` directory into `~/.codex/agents/` for personal use or `<project>/.codex/agents/` for one project. Keep unrelated local profiles and compare same-named files before replacing them.
| Role | Model | Effort | Tier | Purpose |
| --- | --- | --- | --- | --- |
| `Explorer` | `gpt-6-luna` | `medium` | `fast` | Locate primary files in a large codebase when the parent requests it. |
| `Executor` | `gpt-6-luna` | `xhigh` | inherited | Complete a bounded implementation. |
| `Reviewer` | `gpt-6.1-sol` | `high` | inherited | Review a stable result and save a Markdown report when useful. |
| `ExpertAdvisor` | selected per consultation | selected per consultation | inherited | Investigate a complex decision or carry out assigned modeling or automation. |
`ExpertAdvisor.toml` deliberately omits `model` and `model_reasoning_effort`. Choose them at spawn time. Current Codex applies a custom profile's fixed model and effort *after* resolving explicit spawn values, so a fixed profile wins over a conflicting spawn request. Other models and roles may be configured; Team Mode does not impose a global model allowlist. See the [Codex subagent documentation](https://learn.chatgpt.com/docs/agent-configuration/subagents#custom-agents).
## Optional Default Sentinel
Codex includes a built-in general-purpose `default` agent. The repository also provides an optional `default.toml` sentinel, using `gpt-6-luna` at `low` effort. Installing it under `~/.codex/agents/` overrides the built-in default for **all** local tasks and makes an omitted or `default` dispatch return `DISPATCH BLOCKED`. This catches routing mistakes but still creates a child and consumes usage; it is not a security boundary. Team Mode works without it and still asks the parent to name the intended role explicitly.
To disable an installed personal sentinel without deleting it:
```bash
mkdir -p ~/.codex/agents-disabled
mv ~/.codex/agents/default.toml ~/.codex/agents-disabled/default.toml
```
Move it back to `~/.codex/agents/default.toml` to restore it. Open a new task or restart Codex after either change. A project-scoped sentinel can be moved into a corresponding project `agents-disabled/` directory.
## Load And Verify
1. Parse the installed TOML files and check `name`, `description`, `developer_instructions`, model, and effort where configured.
2. Open a new Codex task or restart if role changes do not appear in the current task. The current task's role schema may have been loaded before the edit.
3. Check that `spawn_agent` exposes the intended `agent_type` and, for `ExpertAdvisor`, explicit `model` and `reasoning_effort`. Do not route to a generic child when the intended custom role is absent.
4. For a real delegated task, compare the child trace's role, model, and effort with the request. A file's existence and the child self-report are not runtime proof.
## Explorer Fast Tier
Fast is a service tier, separate from model and reasoning effort. `Explorer.toml` requests it with `service_tier = "fast"`, even when its parent uses Standard. The other profiles inherit their parent's tier. Codex applies this field to the child configuration when Fast mode is enabled. The current `spawn_agent` interface has no per-dispatch service-tier argument. Check the child's effective tier before relying on it because availability depends on the model and host. GPT-6 Fast consumes 2.5 times the Standard credits where available. See the [Speed guide](https://learn.chatgpt.com/docs/agent-configuration/speed), [custom-agent configuration](https://learn.chatgpt.com/docs/agent-configuration/subagents#custom-agents), and [role override implementation](https://github.com/openai/codex/blob/main/codex-rs/core/src/agent/role.rs).
If the runtime does not expose the necessary dispatch controls or ignores a requested role, use the main agent while diagnosing the issue. Restart or update Codex, then check a new task. Do not reuse a version-specific workaround without matching evidence.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |