How does the scientific method apply to AI delegation?

The framework structures prompts so sub-agents move through Observation, Hypothesis, Prediction, Experimentation, Verification, and Conclusion, rather than just executing a list of static commands.

Why shouldn't the orchestrator pre-gather data for sub-agents?

Pre-gathering data (like running lints or tests before delegating) leads to 'context rot' and wastes the orchestrator's tokens. Specialist agents are better equipped to gather their own fresh, comprehensive data using their specialized tools.

What is 'World-Building' in the context of AI resources?

World-building involves describing the development ecosystem—such as authenticated CLIs, available MCP servers, and project-specific scripts—to empower the agent to choose the best tool for the job instead of restricting them to a tiny subset of commands.

What is the core principle of this Agent Orchestration skill?

The core principle is to provide comprehensive 'world-building' context (the Where, What, and Why) and define clear success criteria, while trusting the sub-agent's domain expertise to handle the implementation (the How).

Agent Orchestration

Name: Agent Orchestration
Author: BbgnsurfTech

byBbgnsurfTech

•

생산성 및 워크플로

Provides a scientific delegation framework for orchestrator AIs to coordinate specialist sub-agents while preserving implementation autonomy.

This skill establishes a rigorous methodology for orchestrator AI models to delegate tasks to specialist sub-agents within the Claude Code environment. By focusing on 'world-building' context—defining the Where, What, and Why—instead of prescriptive implementation (the How), it empowers sub-agents to leverage their full 200k context window and specialized toolsets effectively. The framework includes standardized delegation templates, pre-delegation checklists to prevent context rot, and patterns for defining success criteria that align with the scientific method, ensuring high-quality, verifiable outcomes in complex multi-agent workflows.

주요 기능

01Scientific method alignment for agent hypothesis and experimentation

02World-building resource patterns for high-fidelity tool guidance

033 GitHub stars

04Standardized Task tool invocation templates for sub-agent prompts

05Pre-delegation verification checklists to ensure context clarity

06Strategic boundaries that preserve sub-agent autonomy and expertise

사용 사례

01Standardizing task delegation for specialist sub-agents in large repos

02Managing investigative tasks where root causes are unknown

03Coordinating complex multi-agent coding workflows and refactors

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bbgnsurftech/claude-skills-collection agent-orchestration

For use in Claude.ai and ChatGPT

Download Skill