skills.rest
NeoLabHQ
NeoLabHQOFFICIAL
Explore all skills by this author →
WebsiteGitHub

do-competitively

Orchestrates competitive multi-agent generation, multi-judge evaluation, and evidence-based synthesis for tasks.

1.5k|154|Updated Nov 13, 2025
Search Tags:#evaluation#synthesis#llm-judge# orchestration#code design#competitive generation#Multi-Agent
One-click install
npx skills add https://github.com/NeoLabHQ/context-engineering-kit --skill do-competitively
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: do-competitively
Source: https://github.com/NeoLabHQ/context-engineering-kit/tree/main/plugins/sadd/skills/do-competitively
Command: npx skills add https://github.com/NeoLabHQ/context-engineering-kit --skill do-competitively

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Single-pass AI generation often produces suboptimal results for high-stakes tasks like API design, architecture decisions, or complex implementations. This Skill runs multiple independent agents in competition, has them judged against tailored criteria, and synthesizes the best elements into a superior final solution.

Core Features & Use Cases

  • Competitive Generation: Launches 3 independent generator agents plus a meta-judge in parallel, each producing a complete solution with self-critique loops.
  • Multi-Judge Evaluation: 3 independent judges score all solutions against a meta-judge-generated evaluation specification with verification loops.
  • Adaptive Strategy Selection: Automatically chooses SELECT_AND_POLISH for unanimous winners, REDESIGN when all solutions score below 3.0, or FULL_SYNTHESIS for split decisions.
  • Use Case: Design a REST API for user management by having three agents propose different designs (resource-based, RPC-style, minimal), three judges evaluate them, and the winning design gets polished with the best elements from runners-up.

Quick Start

Ask the agent to run do-competitively with a task description such as designing a caching strategy, specifying an output path and evaluation criteria.

Frequently Asked Questions about do-competitively

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

FAQPage Schema
How do I run competitive multi-agent generation for a task?▼

Invoke do-competitively with a task description, output path, and optional evaluation criteria. It dispatches 3 generator agents and 1 meta-judge in parallel, then 3 judges evaluate the solutions, and an adaptive strategy produces the final result.

What is the Generate-Critique-Synthesize pattern?▼

Generate-Critique-Synthesize (GCS) is a multi-phase orchestration where independent agents generate competing solutions with self-critique, judges evaluate them against tailored rubrics, and a synthesizer combines the best elements into a final solution.

When should I use competitive generation instead of a single agent?▼

Use it for high-stakes, well-defined tasks like API design or architecture decisions where quality matters more than speed and cost. Avoid it for trivial tasks, since it runs 8 or more agents and the overhead is not justified.

How does the adaptive strategy selection work?▼

The orchestrator parses judge votes and scores. Unanimous votes trigger SELECT_AND_POLISH, all averages below 3.0 trigger REDESIGN with a return to generation, and split decisions with scores at or above 3.0 trigger FULL_SYNTHESIS.

What are the limitations of multi-agent competitive evaluation?▼

It requires well-defined tasks with clear constraints and rich context; vague descriptions produce incomparable solutions. It also consumes significantly more tokens and time than single-agent execution, making it unsuitable for simple tasks.

Related Skills

Explore More Software Engineering Skills→
thanhquan3010thanhquan3010

iterative-retrieval

Refine codebase context retrieval through iterative query evaluation loops.

Community
Advanced
SchurupIK72SchurupIK72

merge-helper

Automate feature branch integration with validation and documentation updates.

Community
Advanced
mikemylmikemyl
1

go-verified-development

Automate Go code verification with linting, testing, and security scanning tools.

Community
Advanced
melikhanmutlumelikhanmutlu

llm-application-dev-prompt-optimize

Optimize LLM prompts using constitutional AI and chain-of-thought reasoning.

Community
Advanced