Codeagent Wrapper

What it does

Provides a wrapper around codeagent-wrapper that lets agents execute complex code tasks across multiple backends (Codex, Claude, Gemini, OpenCode). It supports agent presets, skill injection, worktree isolation, parallel task orchestration, and structured JSON output — useful for large-scale refactors, multi-agent pipelines, and automated code generation.

When to use it

Use this skill when you need to: deep-analyze codebases, run coordinated multi-step engineering workflows (explore → design → implement → review), execute parallel subtasks with dependency management, or run reproducible CLI-driven agent presets. Prefer --parallel for multi-task runs and --worktree when making filesystem changes.

What's included

• Scripts: none bundled in the skill directory (has_scripts=false).
• References: none in the repo (has_references=false).
• Instructions: CLI examples for single and parallel tasks, agent preset conventions, backend selection, progress frame handling, and structured JSON output formats. Guidance on safe worktree use and rules for long-running tasks is included in the body.

Compatible agents

This skill is designed for agents that can run CLI tools and manage multi-backend LLMs (Codex, Claude, Gemini, OpenCode) and fits into multi-agent orchestrators and developer-focused assistants.