blackbox

Delegate coding tasks to the Blackbox AI CLI multi-model agent.

Updated Jun 5, 2026
One-click install
npx skills add https://github.com/xu1713/openhorse --skill blackbox-xu1713
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: blackbox
Source: https://github.com/xu1713/openhorse/tree/main/openhorse/openhorse/optional-skills/autonomous-ai-agents/blackbox
Command: npx skills add https://github.com/xu1713/openhorse --skill blackbox-xu1713

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @blackboxai/cli.

What problem does it solve? Delegating coding tasks to an external AI agent requires managing interactive terminal sessions, long-running background processes, and multi-model evaluation, which is difficult to orchestrate manually from a chat interface. ## Core Features & Use Cases - One-Shot Task Delegation: Run non-interactive coding tasks with blackbox --prompt against any project directory. - Background Execution & Monitoring: Launch long tasks in background mode and track progress with poll, log, and submit process actions. - Multi-Model Judge Workflow: Dispatch the same task to multiple LLMs (Claude, Codex, Gemini, Blackbox Pro) and let the built-in judge select the best implementation. - Use Case: Ask the agent to add JWT authentication to an Express API; Blackbox runs the task across multiple models, checkpoints the result, and you resume later with a follow-up task like adding rate limiting. ## Quick Start Ask the agent to run a Blackbox one-shot task such as adding JWT authentication to your Express API project in its working directory.

Frequently Asked Questions about blackbox

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

FAQPage Schema
How do I run a one-shot coding task with Blackbox CLI?

Run blackbox with the --prompt flag followed by your task description, for example blackbox --prompt 'Add JWT authentication to the Express API'. Set the workdir to your project directory and always use pty=true since the CLI is an interactive terminal app.

How does Blackbox multi-model mode work?

Blackbox dispatches the same task to multiple LLMs such as Claude, Codex, Gemini, and Blackbox Pro, then uses a built-in judge to evaluate the outputs and select the best implementation. Configure multiple providers via blackbox configure to enable this workflow.

What are the prerequisites for using the Blackbox CLI?

You need Node.js 20 or later, the CLI installed via npm install -g @blackboxai/cli, and an API key from app.blackbox.ai/dashboard configured with blackbox configure. Terminal calls must use pty=true or the CLI will hang.

Can I resume a Blackbox task after it finishes?

Yes, Blackbox has built-in checkpoint support. After a task completes it shows a checkpoint tag, and you resume with blackbox --resume-checkpoint 'tag' --prompt 'follow-up task' to continue work in the same context.

Why does the Blackbox CLI hang when run from a terminal tool?

The CLI is an interactive terminal application and hangs without a pseudo-terminal. Always pass pty=true in terminal calls, and for long tasks use background=true so you can monitor progress with process poll and log actions.