How does the adaptive strategy selection save costs?

It analyzes judge consensus to avoid expensive full synthesis when a clear winner exists, choosing a 'Select and Polish' route that reduces token consumption by 15-20%.

Which models are recommended for this skill?

For high-stakes tasks, using reasoning-heavy models like Claude 3.5 Opus for both generators and judges is recommended to ensure maximum rigor and quality.

Can I use this skill with Cursor or Gemini CLI?

Yes, this skill is part of the Context Engineering Kit and is compatible with Claude Code, Cursor, OpenCode, Antigravity, and Gemini CLI.

What is the Generate-Critique-Synthesize (GCS) pattern?

GCS is a workflow where multiple agents independently generate solutions, evaluate them through rigorous critique, and synthesize the strongest elements into a final, superior output.

What happens if the AI judges give low scores to all solutions?

The skill triggers a 'REDESIGN' strategy, which analyzes the failure modes of the initial attempts and restarts the generation process with new constraints and lessons learned.

Competitive Multi-Agent Orchestrator

Name: Competitive Multi-Agent Orchestrator
Author: NeoLabHQ

byNeoLabHQ

•

542

•

개발자 도구

Executes complex tasks using a competitive multi-agent generation and evaluation framework to ensure superior, production-grade results.

The do-competitively skill implements the Generate-Critique-Synthesize (GCS) pattern to solve high-stakes coding and documentation tasks where quality is more important than speed. By launching three independent generation agents followed by three specialized judges, it eliminates individual model bias and identifies the optimal approach through evidence-based synthesis. The skill features an adaptive strategy engine that intelligently decides whether to polish a clear winner, synthesize a hybrid solution, or trigger a total redesign if quality thresholds aren't met, delivering significantly higher reliability and a 15-20% improvement in cost efficiency.

주요 기능

01Evidence-based synthesis that combines the best elements of parallel solutions

02Automated performance reporting and structured evaluation metrics

03Competitive multi-agent generation with autonomous self-critique loops

04Adaptive strategy selection based on consensus and quality scoring

05542 GitHub stars

06Triple-judge evaluation system using Chain-of-Verification (CoVe) logic

사용 사례

01Drafting complex system architectures or API specifications that require diverse perspectives

02Generating high-stakes technical documentation that must be comprehensive and accurate

03Refactoring critical legacy code where error-free implementation is mandatory

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add neolabhq/context-engineering-kit do-competitively

For use in Claude.ai and ChatGPT

Download Skill