coding-agent

Delegates coding tasks to Codex, Claude Code, or Pi agents via background PTY processes.

1|Updated Mar 3, 2026
One-click install
npx skills add https://github.com/Cyandex/ErnOS --skill coding-agent-cyandex
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: coding-agent
Source: https://github.com/Cyandex/ErnOS/tree/main/skills/coding-agent
Command: npx skills add https://github.com/Cyandex/ErnOS --skill coding-agent-cyandex

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Delegating substantial coding work—building features, reviewing PRs, refactoring large codebases—to interactive CLI agents like Codex or Claude Code is error-prone without proper terminal handling, session monitoring, and workspace isolation. This Skill provides the operational playbook for spawning, monitoring, and managing these agents safely. ## Core Features & Use Cases - PTY-Based Agent Execution: Runs Codex, Claude Code, OpenCode, or Pi with pseudo-terminal allocation so interactive CLI output renders correctly and agents do not hang. - Background Session Management: Spawns agents in background mode and monitors them via process actions (log, poll, write, submit, kill) without blocking the conversation. - Parallel Workflows: Supports batch PR reviews and parallel issue fixing using git worktrees, with one agent per branch or PR. - Use Case: You need to fix three GitHub issues simultaneously. Create a git worktree per issue, launch a background Codex agent in each with --yolo, monitor progress with process logs, then push branches and open PRs. ## Quick Start Ask the agent to run Codex in background mode with PTY enabled in your project directory to build the feature you describe, then check progress with the process log action.

Frequently Asked Questions about coding-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run Codex CLI in the background for a coding task?▼

Run Codex with bash using pty:true, background:true, and a workdir pointing at your project. The call returns a sessionId you can monitor with process log and poll actions, and you can send input with write or submit.

How to review a GitHub PR with Codex without touching my main repo?▼

Clone the repository into a temp directory with mktemp, check out the PR with gh pr checkout, then run codex review against the base branch there. Alternatively, use git worktree add to create an isolated review directory and remove it afterward.

Why does Codex or Claude Code hang or produce broken output when run from a script?▼

Coding agents are interactive terminal applications that require a pseudo-terminal. Without pty:true, output loses formatting or the agent hangs waiting for terminal input, so always allocate a PTY when spawning them.

Why does Codex refuse to run in a plain directory?▼

Codex requires a trusted git repository and refuses to run outside one. For scratch work, create a temp directory and run git init before invoking codex exec with your prompt.

When should I not use a coding agent for a task?▼

Skip agent delegation for simple one-line fixes you can edit directly, for reading code where a read tool suffices, and for thread-bound ACP harness requests in chat. Also never spawn agents inside the assistant's own workspace directory.