codex-implementer

Multi-agent code implementation orchestrates parallel execution where Claude analyzes requirements and partitions tasks by file, then multiple Codex agents write, modify, and refactor code simultaneously across your repository, reducing implementation time compared to sequential approaches.

Yes. This skill delegates code writing and refactoring to Codex CLI in full-auto mode while Claude acts as orchestrator, analyzing requirements, decomposing tasks, reviewing work, and enforcing constraints like type safety and testing integration.

Refactoring tasks route to Codex, which modifies and improves existing code while Claude supervises the process. File-level task partitioning ensures isolated, reviewable changes across your codebase with built-in testing and summary reporting.

Single implementation handles one feature; parallel implementation processes multiple files concurrently; incremental implementation executes stages sequentially with Claude review between steps; refactoring implementation improves existing code—all with explicit file assignments and environment setup.

Yes. The skill applies to both backend endpoints (routes, services) and frontend components, supporting varied architectures. Tasks reference existing code patterns, enforce type safety, and include test integration for consistency across your full stack.

Claude analyzes reference files you specify—such as existing route handlers or component structures—and Codex applies those patterns consistently during implementation, ensuring new code aligns with your codebase conventions without manual guidance on each file.