agent-teams
Coordinate multiple Claude Code sessions as a team — lead + teammates with shared task lists, mailbox messaging, and file-lock claiming. Patterns for team sizing, task decomposition, and when to use teams vs sub-agents vs worktrees.
What this skill does
# Agent Teams
Coordinate multiple Claude Code sessions working on the same codebase simultaneously.
## Enable
```bash
export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1
claude # starts as team lead
```
The first session becomes the team lead. Subsequent sessions in the same repo join as teammates.
## Architecture
```text
Team Lead (coordinates, delegates, reviews)
├── Teammate 1 (owns task A, messages lead + peers)
├── Teammate 2 (owns task B, messages lead + peers)
└── Teammate 3 (owns task C, messages lead + peers)
│
└── Shared: task list + mailbox + file locks
```
**Key difference from subagents:** Teammates are full Claude Code sessions. They have their own context window, can use all tools, and message each other directly — not just report back to a parent.
## Team Sizing
| Team Size | Best For |
|-----------|----------|
| 2 | One builds, one reviews |
| 3-5 | Parallel features across layers (API, UI, tests) |
| > 5 | Coordination overhead outweighs parallelism |
3-5 teammates is the productive range. Beyond that, the lead spends more time coordinating than the team saves.
**Task granularity:** Aim for 5-6 tasks per teammate. Fewer means underutilization; more means excessive context switching.
## Display Modes
**In-process navigation:**
- `Shift+Down` — cycle through teammates (wraps around)
- See each teammate's current task and output
**Split-pane (recommended for >2 teammates):**
- tmux: `tmux split-window -h` per teammate
- iTerm2: Cmd+D for vertical split
- Each pane runs its own `claude` session
## Task Management
Tasks flow through states:
```text
pending → in-progress → completed
│
└── blocked (waiting on dependency)
```
### Task Decomposition
Break work into units that:
- Touch non-overlapping files
- Can be verified independently
- Have clear done criteria
**Good decomposition:**
```text
Task 1: Add rate limiting middleware (src/middleware/rate-limit.ts)
Task 2: Add rate limit tests (tests/rate-limit.test.ts)
Task 3: Update API docs for rate limit headers (docs/api.md)
Task 4: Add Redis config for rate limit store (src/config/redis.ts)
```
**Bad decomposition:**
```text
Task 1: Implement rate limiting
Task 2: Fix rate limiting bugs
Task 3: Improve rate limiting
```
### Dependencies
Tasks can declare dependencies:
```text
Task 3 (API docs) → depends on Task 1 (middleware)
Task 2 (tests) → depends on Task 1 (middleware)
Task 4 (Redis config) → no dependencies
```
Teammates pick up unblocked tasks automatically.
## File-Lock Claiming
Teammates claim files before editing to prevent conflicts:
1. Teammate checks if file is locked
2. If free, claims it (file-lock-based)
3. Edits the file
4. Releases lock on task completion
If two teammates need the same file, one waits or the lead reassigns.
## Plan Approval
Teammates plan before implementing:
1. Teammate receives task
2. Writes a brief plan (files to change, approach)
3. Lead reviews plan
4. Lead approves or redirects
5. Teammate implements
This prevents wasted work from misunderstood requirements.
## Delegate Mode
`Shift+Tab` toggles delegate mode for the lead:
- Lead coordinates only — no direct code edits
- All implementation delegated to teammates
- Lead reviews, approves plans, manages task flow
## Hook Events
| Hook | Fires When |
|------|------------|
| TeammateIdle | A teammate finishes its task and has no pending work |
| TaskCreated | New task added to the shared list |
| TaskCompleted | A teammate marks a task done |
Use these to trigger notifications, auto-assign next tasks, or run integration tests when all tasks complete.
## When to Use Teams vs Alternatives
| Scenario | Use |
|----------|-----|
| Parallel work on non-overlapping files | Agent teams |
| Quick background exploration | Subagent |
| Isolated feature branch work | Worktree (`claude -w`) |
| Competing approaches to same problem | Worktrees (compare results) |
| Cross-layer changes (API + UI + tests) | Agent teams |
| One-off delegation | Subagent with `isolation: worktree` |
| Large-scale migration (20+ files) | `/batch` command |
## Anti-Patterns
- More than 5 teammates — coordination overhead dominates
- Tasks that share mutable state or the same files
- Skipping plan approval — leads to rework
- Lead doing implementation instead of coordinating (use delegate mode)
- Not setting dependencies — teammates race on coupled tasks
## Example Session
```text
Lead: "We need rate limiting. Breaking into 4 tasks."
Task 1: middleware implementation → Teammate 1
Task 2: Redis config → Teammate 2 (no deps)
Task 3: tests → Teammate 1 (after Task 1)
Task 4: API docs → Teammate 2 (after Task 1)
Teammate 1 plans → Lead approves → implements Task 1
Teammate 2 plans → Lead approves → implements Task 2
Task 1 completes → unblocks Tasks 3 and 4
Teammate 1 picks up Task 3, Teammate 2 picks up Task 4
All complete → Lead runs integration tests → commits
```
## Docs
https://code.claude.com/docs/agent-teams
Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.