How do I install the Coding Agents skill?

Run npx skillmds@latest add jorinyang/coding-agents in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Coding Agents safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Coding Agents?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Coding Agents free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under MIT.

Who published Coding Agents?

jorinyang (@jorinyang) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Coding Agents by jorinyang — Agent Skill

Name: Coding Agents
Author: jorinyang

File contents

Coding Agent Delegation

Delegate coding tasks to external AI coding agent CLIs through the Hermes terminal. Each agent has different strengths, auth methods, and operational modes.

Provider Comparison

Feature	Claude Code	OpenAI Codex	OpenCode
Provider	Anthropic	OpenAI	Multi-provider
Install	`npm i -g @anthropic-ai/claude-code`	`npm i -g @openai/codex`	`npm i -g opencode-ai`
Auth	`claude auth login` / `ANTHROPIC_API_KEY`	Codex OAuth / `OPENAI_API_KEY`	`opencode auth login`
One-shot	`claude -p "task"` ✅	`codex exec "task"` ✅	`opencode run "task"` ✅
Interactive	TUI (requires tmux)	TUI (requires PTY)	TUI (requires PTY)
Structured output	JSON schema (`--output-format json`)	No	No
PTY needed	No for `-p`, yes for interactive	Yes	No for `run`, yes for TUI

Mode Selection

Print/exec mode (preferred for most tasks):

# Claude Code
claude -p "Add error handling to API calls" --allowedTools "Read,Edit" --max-turns 10

# Codex
codex exec "Add error handling to API calls"

# OpenCode
opencode run "Add error handling to API calls"

Interactive mode (for multi-turn iterative work):

# Start in tmux
tmux new-session -d -s agent -x 140 -y 40
tmux send-keys -t agent 'cd /project && claude' Enter

Provider Deep Dives

Claude Code — richest feature set: JSON output, session continuation, MCP, hooks, agents, teams. Full reference: references/claude-code.md
OpenAI Codex — simplest: codex exec + --full-auto / --yolo flags. Full reference: references/codex.md
OpenCode — provider-agnostic: opencode run + -f for file attachments. Full reference: references/opencode.md

Common Rules

Always set workdir — keep the agent focused on the right project
Use isolated worktrees for parallel tasks — avoid collisions
Monitor long tasks — use process(action="poll"|"log") for background sessions
Clean up — kill tmux sessions and remove temp worktrees when done
Prefer one-shot mode — cleaner than interactive for bounded tasks
Set turn/budget limits — prevents runaway costs in one-shot mode

Pitfalls

Interactive TUI requires PTY/tmux — pty=true or tmux for TUI agents
Codex requires git repo — use mktemp -d && git init for scratch work
Claude's --dangerously-skip-permissions dialog defaults to "No" — send Down+Enter in tmux
OpenCode /exit opens agent selector — use Ctrl+C instead
PATH mismatch — check which opencode / which codex if behavior differs

---name: coding-agentsdescription: Delegate coding tasks to external AI coding agent CLIs: Claude Code, OpenAI Codex, and OpenCode. Provider comparison, mode selection, and orchestration patterns via Hermes terminal.license: MIT---# Coding Agent DelegationDelegate coding tasks to external AI coding agent CLIs through the Hermes terminal. Each agent has different strengths, auth methods, and operational modes.## Provider Comparison| Feature | Claude Code | OpenAI Codex | OpenCode ||---------|------------|-------------|----------|| Provider | Anthropic | OpenAI | Multi-provider || Install | `npm i -g @anthropic-ai/claude-code` | `npm i -g @openai/codex` | `npm i -g opencode-ai` || Auth | `claude auth login` / `ANTHROPIC_API_KEY` | Codex OAuth / `OPENAI_API_KEY` | `opencode auth login` || One-shot | `claude -p "task"` ✅ | `codex exec "task"` ✅ | `opencode run "task"` ✅ || Interactive | TUI (requires tmux) | TUI (requires PTY) | TUI (requires PTY) || Structured output | JSON schema (`--output-format json`) | No | No || PTY needed | No for `-p`, yes for interactive | Yes | No for `run`, yes for TUI |## Mode Selection**Print/exec mode** (preferred for most tasks):```bash# Claude Codeclaude -p "Add error handling to API calls" --allowedTools "Read,Edit" --max-turns 10# Codexcodex exec "Add error handling to API calls"# OpenCodeopencode run "Add error handling to API calls"```**Interactive mode** (for multi-turn iterative work):```bash# Start in tmuxtmux new-session -d -s agent -x 140 -y 40tmux send-keys -t agent 'cd /project && claude' Enter```## Provider Deep Dives- **Claude Code** — richest feature set: JSON output, session continuation, MCP, hooks, agents, teams. Full reference: `references/claude-code.md`- **OpenAI Codex** — simplest: `codex exec` + `--full-auto` / `--yolo` flags. Full reference: `references/codex.md`- **OpenCode** — provider-agnostic: `opencode run` + `-f` for file attachments. Full reference: `references/opencode.md`## Common Rules1. **Always set `workdir`** — keep the agent focused on the right project2. **Use isolated worktrees for parallel tasks** — avoid collisions3. **Monitor long tasks** — use `process(action="poll"|"log")` for background sessions4. **Clean up** — kill tmux sessions and remove temp worktrees when done5. **Prefer one-shot mode** — cleaner than interactive for bounded tasks6. **Set turn/budget limits** — prevents runaway costs in one-shot mode## Pitfalls1. **Interactive TUI requires PTY/tmux** — `pty=true` or tmux for TUI agents2. **Codex requires git repo** — use `mktemp -d && git init` for scratch work3. **Claude's `--dangerously-skip-permissions` dialog defaults to "No"** — send Down+Enter in tmux4. **OpenCode `/exit` opens agent selector** — use Ctrl+C instead5. **PATH mismatch** — check `which opencode` / `which codex` if behavior differs

Frequently asked questions

Coding Agents