How does this skill detect if a team is needed?

It uses complexity heuristics based on the specific command, such as the number of files changed, whether changes span frontend and backend directories, or if the task involves multi-branch rebasing.

What is the token cost of using Agent Teams?

Agent Teams typically cost 3-5x more tokens than a single agent run, but they are designed to save significant developer time on complex, high-stakes tasks.

When should I use Agent Teams instead of a single agent?

Teams are best for complex tasks like fullstack TDD, large-scale code reviews (>100 lines), or debugging where multiple competing hypotheses need testing to find a root cause.

What is a Claude Code Agent Team?

Agent Teams coordinate multiple independent Claude Code sessions that can communicate and share tasks, allowing for parallel execution and adversarial collaboration unlike standard subagents.

Does this work with standard Claude Code environments?

Yes, but it requires the experimental flag CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 to be enabled to function.

Agent Team Coordinator

Name: Agent Team Coordinator
Author: lenneTech

bylenneTech

0•

협업 도구

Coordinates multiple Claude Code agents using intelligent heuristics and patterns for parallel debugging, reviews, and fullstack development.

The Agent Team Coordinator skill optimizes the experimental Claude Code Agent Teams feature by providing sophisticated auto-detection heuristics and specialized coordination patterns. It determines when parallel agent workflows outperform single-agent execution based on task complexity, such as fullstack TDD, adversarial debugging, or multi-file code reviews. By implementing patterns like 'Independent Then Challenge' and 'Adversarial Convergence,' it ensures that multi-agent workflows are utilized only when the quality improvement or time savings justify the token overhead, enabling true parallelism in complex software engineering tasks.

주요 기능

01Auto-detection heuristics for team vs. single-agent mode selection

020 GitHub stars

03Adversarial debugging patterns for competing hypothesis testing

04Parallel worktree execution for high-speed batch git operations

05Inter-agent communication protocols via direct messaging and broadcasting

06Quality gate hooks to validate teammate output and task deliverables

사용 사례

01Parallelizing fullstack TDD by coordinating backend contract definition with frontend consumption

02Resolving complex bugs through competing hypotheses and active falsification sessions

03Conducting unbiased, independent code reviews across large, multi-file changesets

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lennetech/claude-code coordinating-agent-teams

For use in Claude.ai and ChatGPT

Download Skill