How does this skill prevent session collisions?

The skill uses unique Controller IDs to tag and isolate threads, ensuring that multiple open terminal windows or Codex sessions do not interfere with each other's subagents.

How do I see what a subagent has completed?

You can use the 'peek' command to fetch the newest unseen messages, the 'log' command to view full history, or the 'watch' command for real-time polling updates.

Can I launch multiple tasks at once?

Yes, you can use the manifest-based launching feature to describe several prompts, personas, and tasks in a single JSON file for simultaneous background execution.

What is a codex-subagent?

A codex-subagent is a background helper process launched by Claude Code to handle specific tasks asynchronously, preventing the main conversation from becoming bloated with context.

Does it work across different repositories?

The skill includes best practices for encoding workspace paths in prompts to ensure subagents always operate within the correct directory, regardless of where the CLI is called.

Codex Subagent Management

Name: Codex Subagent Management
Author: synqing

bysynqing

生产力与工作流

Orchestrates background subagents to perform parallel research and long-running tasks without bloating the primary Claude Code session.

关于

This skill provides a comprehensive framework for managing codex-subagents, allowing developers to offload research, context-heavy builds, or complex asynchronous work to background processes. By utilizing disk-based prompt and result handling via the .codex-subagent directory, it ensures that the primary chat remains lean and focused. The skill includes robust controller ID isolation to prevent multi-terminal session collisions, manifest-based batching for complex workflows, and real-time monitoring through specialized peek and watch commands, making it an essential utility for large-scale development projects.

主要功能

Asynchronous background execution for long-running parallel tasks
Real-time monitoring with peek, watch, and wait command status updates
Batch task launching via JSON manifests for complex workflows
Disk-based prompt and result persistence for reproducible history
Controller ID tagging to isolate multiple active Codex sessions
0 GitHub stars

使用场景

Executing automated batch code reviews using specialized personas and manifests
Performing deep background research while continuing main thread development
Managing context-heavy builds or large-scale refactors in parallel threads

关于

主要功能

Asynchronous background execution for long-running parallel tasks
Real-time monitoring with peek, watch, and wait command status updates
Batch task launching via JSON manifests for complex workflows
Disk-based prompt and result persistence for reproducible history
Controller ID tagging to isolate multiple active Codex sessions
0 GitHub stars

使用场景

Executing automated batch code reviews using specialized personas and manifests
Performing deep background research while continuing main thread development
Managing context-heavy builds or large-scale refactors in parallel threads