Skip to content

mikeylong/balanced-team

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 

Repository files navigation

balanced-team

balanced-team is an Agent Skill for execution tasks that need a small coordinated team, a concrete plan, implementation, and a bounded review/fix cycle.

Use it as shorthand for:

Recruit and coordinate a balanced team to handle this task, then plan, implement, and review with up to two review loops where subsequent findings can be discovered and addressed.

Recommendation (July 2026)

Codex: Start with GPT-5.6 Sol where it is available. OpenAI positions Sol as its flagship GPT-5.6 model for complex coding, research, computer use, and other open-ended work. Codex's Ultra mode also uses subagents when a task can benefit from parallel execution.

In my use, Sol handles task decomposition, delegation, integration, and proportional review well from a direct prompt. This skill can add more workflow than the task needs. Try the simpler instruction first:

Use a small team only where the work can proceed independently. Keep one lead responsible for integration, verify the combined result end to end, and run at most two review/fix passes—stopping early when another pass has little expected value.

Anthropic / Claude Code: Claude Code now has native Agent Teams with a lead session, independent teammates, shared tasks, and direct messaging. As of July 2026, Agent Teams are experimental and disabled by default. Anthropic recommends them for independent research, review, debugging, and cross-layer work, while warning that they add coordination overhead and use significantly more tokens than a single session.

The native Claude Code feature covers most of the behavior this skill encodes. balanced-team may still be useful as a portable fallback while Agent Teams are experimental or when you want the same prompt across tools, but it is not required once Agent Teams are enabled. For focused or sequential work, Anthropic recommends subagents or a single session instead.

Given the current native capabilities in Codex and Claude Code, treat this repository as a reference and fallback rather than a default installation.

Install

Install this repository anywhere your skills-compatible agent loads custom skills from. For filesystem-based clients, clone or copy the repository into that client's skills directory:

git clone https://github.com/mikeylong/balanced-team.git /path/to/skills/balanced-team

Restart or reload your agent if it only discovers skills at startup.

Examples

Use the balanced-team skill to add CSV export to the dashboard.
Use balanced-team to review and harden this release plan.

About

Agent Skill for balanced team execution with bounded review loops

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors