skills.rest
ruiyangruiyi
ruiyangruiyiCOMMUNITY
Explore all skills by this author →
GitHub

blackbox

Delegate coding tasks to Blackbox AI CLI with multi-model judging.

Updated May 26, 2026
Search Tags:#code generation#multi-model#PR review#interactive cli#checkpoint resume#terminal execution#delegate coding
One-click install
npx skills add https://github.com/ruiyangruiyi/hermes-agent --skill blackbox-ruiyangruiyi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: blackbox
Source: https://github.com/ruiyangruiyi/hermes-agent/tree/main/optional-skills/autonomous-ai-agents/blackbox
Command: npx skills add https://github.com/ruiyangruiyi/hermes-agent --skill blackbox-ruiyangruiyi

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Coding tasks often require trying multiple approaches and model opinions to get correct, high-quality implementations faster, especially when you need a strong review and selection process.

Core Features & Use Cases

  • Multi-model code generation: Runs the same coding task across multiple LLM backends (e.g., Claude, Codex, Gemini, Blackbox Pro) to increase the chance of a correct solution.
  • Built-in judging to pick the best result: Compares outputs and selects the strongest implementation for you to use.
  • Interactive and background execution: Supports one-shot prompts as well as longer-running tasks via background sessions, with polling and logging.
  • Checkpoints and resume: Lets you pause and resume work using checkpoint tags when tasks are lengthy or iterative.
  • Developer workflow support: Helps with PR review scenarios by delegating review prompts without directly modifying your main working tree.

Quick Start

Delegate a coding change by telling the Hermes agent to run Blackbox in PTY mode in the target project directory with your exact prompt.

Frequently Asked Questions about blackbox

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

FAQPage Schema
How do I compare multiple LLM code generation outputs for the same task?▼

Multi-model code generation works by running the same coding prompt across different LLM backends like Claude, Codex, and Gemini. A built-in judging mechanism then compares the outputs and selects the strongest implementation for you to use.

How do I resume a long-running code refactor or debugging session?▼

You can resume a long-running code refactor using checkpoint tags. This allows you to pause and later resume iterative coding work or interactive debugging sessions without losing your previous progress.

Do I need Node.js to run multi-model coding and PR review tasks?▼

Yes, you need Node.js 20+ and the blackbox CLI with a configured API key to run multi-model coding and PR review tasks. The execution uses PTY mode to support interactive terminal workflows.

Can I delegate PR review prompts without modifying my working tree?▼

Yes, you can delegate PR review prompts without modifying your main working tree. The tool supports developer workflows by running review sessions in interactive or background execution modes.

What is the best way to run interactive coding sessions for adding tests?▼

The best way to run interactive coding sessions for adding tests is via PTY execution with background polling and logging. This supports terminal workflows and allows the judge to select the best implementation.

Does multi-model code generation support background execution?▼

Yes, multi-model code generation supports background execution for lengthy tasks. It utilizes background sessions with polling and logging to manage one-shot prompts and long-running refactors.

Related Skills

Explore More Software Engineering Skills→
nadicodeainadicodeai

blackbox

Delegate coding tasks to multiple AI models via Blackbox CLI.

Official
Intermediate
rongarederongarede
4

collaborating-with-codex

Delegate coding tasks to Codex CLI via a Python bridge with JSON outputs.

Community
Intermediate
JKhyroJKhyro

codex

Delegate coding tasks to the Codex CLI from Hermes terminal sessions.

Community
Advanced
1thirteeng31thirteeng3

blackbox

Delegate coding tasks to multiple AI models with built-in judging.

Community
Advanced