What problem does it solve?
This Skill provides a reliable way to run and manage interactive coding agents (Codex, Claude Code, OpenCode, Pi) from the shell while ensuring correct pseudo-terminal allocation, workspace isolation, and background session control so agents do not hang or read unrelated files.
Core Features & Use Cases
- PTY-required execution: Ensures interactive CLIs receive a pseudo-terminal to avoid broken output, missing colors, or hangs.
- Background sessions & monitoring: Start agents in background, receive session IDs, stream logs, poll status, send input, paste text, and terminate sessions as needed.
- Tool-specific guidance and safety: Covers Codex git-repo requirements and flags, Claude Code/OpenCode/Pi usage, safe PR review patterns with temporary clones or worktrees, and auto-notify on completion for long tasks.
- Use Case: Launch parallel Codex reviewers across many PRs in isolated worktrees, monitor each session, and post aggregated results to GitHub.
Quick Start
Start an interactive coding agent in your project directory as a background PTY session and monitor progress with process action:log.