Skip to content

tobihagemann/turbo / claude/skills/codex-exec

codex-exec

Verified

Run autonomous task execution using the codex CLI. Use when the user asks to "codex exec", "run codex exec", "execute a task with codex", or "delegate to codex".

FileKindSize
SKILL.mdskill md10.7 KB
references/parallel-execution.mdmarkdown1.8 KB

references/parallel-execution.md

View on GitHub
# Parallel Execution in Codex

Codex supports parallel sub-agents via `spawn_agent` and `wait_agent` tools. This works in both interactive and `codex exec` mode.

## How It Works

- **`spawn_agent`** creates a new agent thread with its own conversation context. Returns an agent ID.
- **`wait_agent`** waits for one or more agents to complete and returns their results.
- Sub-agents run concurrently, not blocking the parent.
- The parent can spawn multiple agents, do its own work, then wait for results.

## Gating Rule

The model will not fan out into sub-agents unless the prompt **explicitly requests parallel delegation**. The `spawn_agent` tool description gates itself on explicit user permission:

> "Only use spawn_agent if and only if the user explicitly asks for sub-agents, delegation, or parallel agent work."

Prompts must clearly say to use sub-agents or parallel execution.

## Prompt Pattern

To trigger parallel fan-out, structure the prompt like:

```
Delegate each of the following tasks to a separate sub-agent using spawn_agent so they run in parallel. Wait for all agents to complete, then synthesize their findings into a unified report.

1. [Task A description]
2. [Task B description]
3. [Task C description]
```

Key elements:
- Explicitly mention `spawn_agent` or "sub-agent" or "parallel"
- List the tasks to fan out
- Instruct to wait for all results and synthesize

## Limitations

- **Thread limits**: `agents.max_threads` in codex config caps concurrent sub-agents (Codex default 6; Turbo's `codex/SETUP.md` recommends raising it to 16). The hard ceiling enforced by Codex is 64.
- **Depth limits**: `agents.max_depth` (default 1) prevents unbounded recursion — a sub-agent cannot itself spawn sub-agents under the default config
- **No structured output**: Sub-agents return text via `wait_agent` / `resume_agent` results; there is no schema-typed channel
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: 471kUpdated Claude CodeCodexUnverifiedInstall
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: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
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 availableStars: 322kUpdated Claude CodeCodexVerifiedInstall