When should I choose Agent Teams over Subagents?

Choose Agent Teams when the task requires inter-agent discussion, adversarial refinement, or multi-perspective debate. If the workers need to talk to each other to reach a conclusion, a team is necessary.

What are the best use cases for Task() subagents?

Subagents are ideal for 'fire-and-forget' tasks where only the result matters, such as running quality gates, linting, single-file implementations, or broad information gathering.

How does Tiered Delegation reduce token costs?

It provides clear thresholds for when to use cheaper 'single session' or 'subagent' methods, preventing the high token overhead of a 3-person team for routine tasks that don't benefit from collaboration.

What is the recommended escalation path for complex work?

The framework suggests starting with the cheapest method (subagents). You should only escalate to persistent teams if the results require synthesis or debate that the orchestrator cannot handle alone.

Tiered Delegation

Name: Tiered Delegation
Author: rbergman

byrbergman

•

协作工具

Optimizes task execution by providing a decision framework to choose between single sessions, subagents, and collaborative agent teams.

The Tiered Delegation skill provides a strategic framework for managing Claude's workload by categorizing tasks into three distinct tiers based on complexity and coordination needs. It helps developers navigate the trade-offs between speed, cost, and depth of insight by identifying which tasks can be handled in a single session, which require the 'fire-and-forget' efficiency of subagents, and which demand the adversarial refinement of persistent agent teams. By following this structured escalation path, users can minimize token overhead while ensuring that complex architectural decisions and multi-module implementations receive the necessary multi-perspective deliberation.

主要功能

01Structured escalation paths from simple execution to complex team debate

02Cost-aware decision logic to minimize unnecessary token consumption

03Three-tier coordination framework for granular task management

04Sequential decision questions to determine optimal agent interaction levels

052 GitHub stars

06Specific 'Sweet Spot' criteria for subagents versus persistent teams

使用场景

01Managing complex multi-module implementations that require shared interface coordination

02Determining if a code review needs parallel subagents or a collaborative debating team

03Optimizing research tasks by separating data gathering from adversarial hypothesis testing

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rbergman/dark-matter-marketplace tiered-delegation

For use in Claude.ai and ChatGPT

主要功能

01Structured escalation paths from simple execution to complex team debate

02Cost-aware decision logic to minimize unnecessary token consumption

03Three-tier coordination framework for granular task management

04Sequential decision questions to determine optimal agent interaction levels

052 GitHub stars

06Specific 'Sweet Spot' criteria for subagents versus persistent teams

使用场景

01Managing complex multi-module implementations that require shared interface coordination

02Determining if a code review needs parallel subagents or a collaborative debating team

03Optimizing research tasks by separating data gathering from adversarial hypothesis testing