XNews

Claude Code / Codex Ecosystem and Agent Tooling

Claude Code / Codex Ecosystem and Agent Tooling

OpenAI announced GPT-5.2-Codex as an agentic coding model for complex long tasks, improved security, and visual UI understanding; Codex now supports Skills and multiple ecosystem integrations.

Key insights

GPT-5.2-Codex targets agentic engineering

OpenAI released GPT-5.2-Codex positioned as an 'Agentic Coding Model' that operates in real terminals, writes and debugs code, and sustains long-running task focus. The model reports native long-context handling (context compression), task stability across plan changes, and improved terminal/Windows-native operations. Benchmarks cited include SWE-Bench Pro (56.4%) and Terminal-Bench 2.0 (64.0%).

Codex adds Skills compatible with Claude Skills

OpenAI announced Codex now supports 'Skills'—a reification of instruction sets, scripts and resources in SKILL.md format. Skills can be invoked via commands like $.skill-name or triggered automatically by context. The Skills feature follows Claude's approach to progressive loading: metadata up front, full content on demand, enabling more efficient context use.

Ecosystem: Claude Code compatibility and tooling expansion

OpenRouter and other projects released Claude Code-compatible integrations; third-party models and vendors (Kimi, GLM, Minimax, DeepSeek) also provide compatibility. Claude Chrome plugin expanded to paid users, bringing sidebars, multi-tab workflows, console log reading, and support for running/testing code in-browser—integrated with Claude Code capabilities.

Source posts