kimi-supervisor

Delegate research to Kimi K2.5 and verification to MiniMax for token-efficient workflows.

Updated Dec 15, 2025

Search Tags:#research #multi-agent #orchestration #token efficiency #cost optimization #codebase exploration

Ready to run

npx skills add https://github.com/OpenSourceSam/v2_heras_garden --skill kimi-supervisor

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: kimi-supervisor
Source: https://github.com/OpenSourceSam/v2_heras_garden/tree/main/.claude/skills/kimi-supervisor
Command: npx skills add https://github.com/OpenSourceSam/v2_heras_garden --skill kimi-supervisor

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill drastically reduces the token costs associated with complex research and exploration tasks by offloading the bulk of the work to more cost-effective AI models like Kimi and MiniMax, while Claude retains strategic control.

Core Features & Use Cases

Token Savings: Achieves 85-93% token savings on exploration tasks compared to direct Claude usage.
Multi-Agent Orchestration: Claude delegates research to Kimi, which synthesizes findings, and then uses MiniMax for verification before Claude consumes the summarized output.
Use Case: When Claude needs to research a large codebase for specific patterns or information, it can use Kimi to perform the deep dive and MiniMax to validate the findings, providing Claude with a concise, verified summary instead of raw, expensive output.

Quick Start

Use the kimi-supervisor skill to research codebase patterns by having Claude delegate tasks to Kimi for research and MiniMax for review.

Frequently Asked Questions about kimi-supervisor

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

FAQPage Schema

How do I reduce token consumption during complex codebase exploration with Claude?▼

Multi-agent orchestration reduces token consumption by delegating exploration tasks to Kimi K2.5 and verification to MiniMax. Claude receives concise, verified summaries instead of raw output, achieving 85-93% token savings on research workflows.

What is multi-agent orchestration for AI research workflows?▼

Multi-agent orchestration is a workflow where Claude delegates deep research to Kimi, which synthesizes findings, and MiniMax verifies the results. This provides Claude with a concise, validated summary instead of processing expensive raw output directly.

Do I need specific environment variables to use Kimi and MiniMax for delegated research?▼

Yes, multi-agent research requires specific environment variable setup for Kimi and MiniMax agents. These configurations allow Claude to securely delegate exploration and verification tasks to the respective models.

Can I perform batch operations and information synthesis across multiple domains?▼

Yes, the orchestration supports complex research across multiple domains, including batch operations and information synthesis. Kimi handles the diverse source exploration, and MiniMax validates the synthesized findings before returning them.

What is the best way to verify AI research findings without high token costs?▼

The best way to verify findings cost-effectively is delegating validation to MiniMax after Kimi performs the initial research. Claude then reviews the verified summary, ensuring accuracy while maintaining 85-93% token savings.