coding-agent-cli

Delegate coding tasks to external CLI coding agents via terminal orchestration.

10|2|Updated May 6, 2026
One-click install
npx skills add https://github.com/jcrabapple/hermes-skills --skill coding-agent-cli-jcrabapple
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: coding-agent-cli
Source: https://github.com/jcrabapple/hermes-skills/tree/main/coding-agent-cli
Command: npx skills add https://github.com/jcrabapple/hermes-skills --skill coding-agent-cli-jcrabapple

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Delegating feature work, refactoring, PR reviews, and batch issue fixing to headless coding agents requires knowing each CLI's install, auth, flags, and quirks. This Skill consolidates orchestration recipes for Claude Code, Codex, OpenCode, and Pi into one playbook. ## Core Features & Use Cases - Two orchestration modes: one-shot print mode for bounded tasks and interactive PTY via tmux for multi-turn sessions with monitoring and follow-up input. - Per-agent references: detailed guides covering Claude Code flags and hooks, Codex sandbox modes, OpenCode TUI keybindings, and Pi provider configuration. - Parallel workers and guardrails: run multiple agents in isolated worktrees with --max-turns, --max-budget-usd, and --allowedTools to control cost and permissions. - Use Case: Fan out three GitHub issues to parallel Claude Code instances in tmux sessions, each fixing one issue with restricted tools, then review the diffs before merging. ## Quick Start Ask the agent to delegate a task such as fixing a bug in src/auth.py to Claude Code in print mode with a turn limit and restricted tools.

Frequently Asked Questions about coding-agent-cli

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

FAQPage Schema
How do I delegate a coding task to Claude Code from the terminal?▼

Use print mode for one-shot tasks: claude -p 'your task' with --allowedTools to restrict capabilities and --max-turns to bound the agentic loop. Print mode skips workspace trust and permission dialogs, making it suitable for automation.

Claude Code vs Codex vs OpenCode: which coding CLI should I use?▼

Claude Code is the default with structured JSON output and --from-pr for reviews. Codex fits users with OpenAI OAuth but requires pty=true and a git repo. OpenCode is provider-agnostic, accepting OpenRouter, Anthropic, or Ollama keys.

How do I run an interactive coding agent session with tmux?▼

Create a detached tmux session, launch the agent inside it with send-keys, then monitor with capture-pane and send follow-up prompts with send-keys. Always kill the tmux session when finished to avoid resource leaks.

Why does Codex CLI hang or fail when run from a script?▼

Codex is an interactive terminal app and hangs without pty=true in the terminal call. It also refuses to run outside a git repository, so use mktemp -d && git init for scratch work.

How do I prevent runaway costs when using headless coding agents?▼

Set --max-turns and --max-budget-usd in print mode to cap agentic loops and spend. Restrict capabilities with --allowedTools such as Read,Edit for reviews, and set a timeout of at least 120 seconds on the surrounding terminal call.

When should I not use an external coding agent CLI?▼

Avoid external CLIs for chatting about code, small inline edits, or quick yes/no questions on changes; use direct file reads, patches, or the agent's own reasoning instead. All four CLIs also require network access to their model providers.