skills.rest
heavy3-ai
heavy3-aiOFFICIAL
Explore all skills by this author →
WebsiteGitHub

h3

Automates multi-model code review feedback using OpenRouter with web search and 200K token context.

45|15|Updated Jan 24, 2026
Search Tags:#multi-model#council#diff#PR#code-review#OpenRouter#web-search
One-click install
npx skills add https://github.com/heavy3-ai/code-audit --skill h3
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: h3
Source: https://github.com/heavy3-ai/code-audit/tree/main/skill
Command: npx skills add https://github.com/heavy3-ai/code-audit --skill h3

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires requests, and includes scripts (resource) components.

What problem does it solve?

Heavy3 Code Audit automates multi-model code reviews via OpenRouter to identify bugs, edge cases, and security concerns in code changes. It speeds up review cycles by aggregating insights from multiple models and delivering structured feedback to developers.

Core Features & Use Cases

  • Single-model Review: fast, cost-effective feedback using Kimi K2.5
  • Council Mode: 3-model analysis with GPT 5.2 + Gemini 3 Pro + Grok 4
  • Web Search Integration: surface external references during reviews
  • Large Context Support: up to 200K token context for big diffs and plans

Quick Start

Use the h3 skill to review a PR: /h3 pr <number> or review a plan file: /h3 plan.md or review a Markdown file: /h3 <file>.md You can switch to council mode with --council or use --free for a rotating free model.

Frequently Asked Questions about h3

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

FAQPage Schema
How do I automate multi-model code review for a PR?▼

You can automate multi-model code review for a PR by running the h3 skill with a PR number. It collects diffs and documentation via OpenRouter to deliver structured feedback and identify bugs.

Can I use a three-model council for automated code audit?▼

Yes, you can use a three-model council for automated code audit by enabling the --council flag. It analyzes code changes using GPT 5.2, Gemini 3 Pro, and Grok 4 to surface edge cases and security concerns.

Does multi-model code review support large context diffs?▼

Multi-model code review supports large context diffs up to 200K tokens. This allows you to process big plans and extensive code changes without truncating the input context.

How do I integrate web search into an automated code review?▼

You integrate web search into automated code review by using the council mode in h3. It surfaces external references during the analysis to provide actionable insights and validate code changes.

Do I need an OpenRouter API key to run multi-model code review?▼

Yes, you need an OpenRouter API key to run multi-model code review. The local configuration uses this key to access models, collect diffs, plan content, and generate feedback.

What is the best way to get fast code review feedback on a budget?▼

The best way to get fast code review feedback on a budget is using single-model mode with Kimi K2.5 or the --free flag for a rotating free model, delivering cost-effective insights.

Related Skills

Explore More Software Engineering Skills→
HaykTarkhanyanHaykTarkhanyan

perplexity-search

Automate web searches with Perplexity models via OpenRouter and citations.

Community
Advanced
robbytrobbyt
42

diff-review

Review git diffs for bugs, security issues, and style problems.

Community
Advanced
jaminitachijaminitachi
1

automation-pipeline

Design, build, and execute multi-step automation pipelines with configurable triggers.

Community
Advanced
ohyesgocoolohyesgocool

mr-review

Orchestrate independent AI model code reviews on GitLab or GitHub merge requests.

Community
Advanced