codex-claude-loop

Dual-AI orchestration coordinates Claude for planning and implementation with Codex for validation and review. Set up a .codex-loop directory, configure git status checks, and use codex exec commands to automate the iterative feedback loop across planning, implementation, and code review phases in non-TTY environments.

Use AI orchestration to separate concerns: have Claude handle architecture and code implementation, then route the output to Codex for rigorous validation of architectural flaws, security vulnerabilities, and performance bottlenecks. This structured feedback loop minimizes errors and guides iterative fixes until quality standards are met.

Yes. The dual-AI engineering loop runs in non-TTY environments using codex-cli with timeout control and command workflows. It automates pre-flight checks, plan validation, code review, and iterative fixes, producing documented artifacts at each phase suitable for pipeline integration.

Role separation ensures Claude focuses on architecture and implementation expertise while Codex provides independent validation and review. This division prevents single-point-of-failure reasoning, catches interdependent issues earlier, and enforces quality gates throughout the development cycle.

Use dual-AI orchestration for complex feature development, large-scale refactoring, and projects with stringent security or performance requirements. The iterative loop is ideal when functional and technical validation must happen in separate specialized phases to meet high quality standards.

Create a .codex-loop directory for workflow state, ensure git status tracking is configured, and install codex-cli. The skill operates in non-TTY environments with timeout control, automating multi-phase engineering loops from planning through documentation without interactive terminal input.