Codex Agent Delegate

0•

Delegates complex coding, browser research, and code review tasks to independent Codex CLI sessions for enhanced multi-tasking.

The Codex Agent skill empowers Claude to orchestrate specialized sub-agents by interfacing directly with the Codex CLI. It allows for the delegation of intensive tasks such as large-scale refactors, deep code reviews, and autonomous browser-based research using 'Computer Use' capabilities. By offloading these workflows to standalone Codex sessions, users can maintain cleaner primary contexts while benefiting from multi-turn session persistence, structured JSON outputs, and secure execution environments via configurable sandbox modes.

主要功能

01Structured output support with JSON Schema validation and file logging
02Secure execution via configurable read-only or workspace-write sandboxes
03Autonomous browser research and 'Computer Use' functionality
04Multi-turn session persistence with thread_id tracking and resumption
05Advanced code review sub-commands for git-based analysis
060 GitHub stars

使用场景

01Executing complex multi-file code refactors in a dedicated, resumable agent environment
02Automating deep code reviews against specific git branches or uncommitted changes
03Performing autonomous web research and summarizing findings via a headless browser session