agentleFS
Sign inStart free
sergeeeysergeeey/Claude-cod-top-2026/skills/core/agent-teams/SKILL.md

agent-teams

Skill9 starsChanged 6 months agoMITView on GitHub

Паттерны для оптимального состава команд агентов: sizing heuristics, preset configurations, agent type selection. Используй когда решаешь сколько агентов запустить, какой тип агента выбрать для роли, как организовать Review/Debug/Feature/Security команду. Триггеры: /agent-teams, "team composition", "how many agents", "spawn agents", "multi-agent team", "сколько агентов", "команда агентов", "parallel agents setup".

16 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use codex-review.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • juriwiensSkill

    codex-review

    Run code review via Codex CLI

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 5 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When team agents disagree:

    line 96
  2. Distinctive:

    If nearing budget: lead summarizes and stops teammates.

    line 109
  3. Distinctive:

    DON'T: use teams for trivial tasks (single file, obvious fix)

    line 113
  4. Distinctive:

    DON'T: run build-squad without a clear spec (both agents need the same input)

    line 114
  5. Distinctive:

    DON'T: run review-squad on MVP code — single reviewer is enough

    line 116

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md
<!-- BSV — Brief Skill View | поиск: BSV Скил : agent-teams TL;DR : Sizing heuristics + preset teams + agent type selection для multi-agent workflows Вызов : /agent-teams, team composition, how many agents, spawn agents, multi-agent НЕ для : Одиночный агент (прямо используй Agent tool); review кода (→ /reviewer) -->

Agent Teams Orchestration

Sizing Heuristics

Правило: Начни с наименьшей команды которая покрывает все нужные измерения. Каждый дополнительный агент = overhead координации.

СложностьРазмерКогда использовать
Simple1-2Single-dimension review, изолированный баг, небольшая фича
Moderate2-3Multi-file изменения, 2-3 аспекта, средние фичи
Complex3-4Cross-cutting concerns, большие фичи, глубокий debug
Very Complex4-5Full-stack фичи, comprehensive review, системные проблемы

Agent Type Selection

ТипИнструментыИспользовать для
general-purposeВсеImplementation, debugging
ExploreRead-onlyResearch — НЕ для impl
PlanRead-onlyАрхитектурное планирование
reviewerRead + BashCode review
builderRead + Write + Edit + BashНаписание кода
testerRead + Write + BashНаписание тестов

Критическое: Read-only агенты (Explore, Plan) НЕ могут изменять файлы.

Preset Team Compositions

КомандаАгентыИзмеренияКогда
Debug Team (3)3x general-purposeПо 1 competing hypothesis каждомуБаг с несколькими причинами
Feature Team (3)1x lead + 2x builderПараллельные workstreamsФича делится на части
Fullstack Team (4)1x lead + frontend + backend + testsFull stackCross-layer фича
Security Team (4)4x reviewerOWASP / auth / dependencies / secretsComprehensive audit
Migration Team (4)1x lead + 2x builder + 1x reviewerParallel impl + correctnessБольшая миграция

Available Teams (наши стандартные)

TeamLeadTeammateStrategyUse Case
review-squadreviewersec-auditorparallelCode review + security audit
build-squadbuildertesterparallel-worktreeImplementation + tests
research-squadexplorerverifiersequentialSearch + verify claims

Decision Matrix: When to Use Teams

SituationSingle AgentTeam
Simple code review (1-2 files)reviewer—
Review touching auth/payment—review-squad
New feature with clear specbuilderbuild-squad
Quick codebase searchexplorer—
Research with Evidence Policy—research-squad
Multi-file refactoring (3+)—review-squad

SendMessage Pattern (Multi-Turn Agent Conversations)

To continue a conversation with a completed subagent:

SendMessage(to: "agent-<id>", message: "Now check edge cases for...")

The agent resumes with full context — no reload, no token waste.

When to use:

  • Explorer found partial results → send follow-up query
  • Builder needs to iterate on reviewer feedback
  • Verifier needs more context to verify a claim

Conflict Resolution

When team agents disagree:

  1. Security wins — if sec-auditor says BLOCKED, the verdict is BLOCKED
  2. Evidence wins — [VERIFIED] claims override [INFERRED]
  3. Lead decides — if ambiguous, lead agent makes the final call

Token Budget Management

TeamTypical CostBudget Cap
review-squad~1500 tok2500 tok
build-squad~2500 tok4000 tok
research-squad~1200 tok2000 tok

If nearing budget: lead summarizes and stops teammates.

Anti-Patterns

  • DON'T use teams for trivial tasks (single file, obvious fix)
  • DON'T run build-squad without a clear spec (both agents need the same input)
  • DON'T ignore verifier's HALLUCINATION verdict — it means the claim has no evidence
  • DON'T run review-squad on MVP code — single reviewer is enough

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free