claude-code-teams
por matthew-a-gordon
Web & Frontend DevelopmentResumen
Production skill for orchestrating Claude Code's native agent teams feature.
Descripción
Qué hace esta skill
Orchestrates multiple Claude Code agents working in parallel on complex tasks. Agents coordinate through a shared task list, message each other directly, and own separate file boundaries to avoid conflicts. Includes templates for common patterns like multi-lens reviews, competing-hypothesis debugging, and full-stack feature builds.
Parallel agent execution cuts wall-clock time for independent workstreams that would otherwise be bottlenecked by a single context window working sequentially.
Cuándo usarla
- Running security, performance, and test reviews simultaneously on a PR
- Debugging a hard error by testing two competing root-cause theories in parallel
- Building a full-stack feature with separate agents for frontend, backend, and tests
- Generating an ADR through adversarial debate between architecture approaches
- Classifying or refactoring a large inventory of files in parallel
Flujo de trabajo de ejemplo
User asks: Review this PR for security issues, performance regressions, and missing test coverage
- 1
Enable CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 in environment
- 2
Select the parallel-review.md template
- 3
Paste the spawn prompt into Claude Code via --pty
- 4
Lead agent assigns security, performance, and test scopes to three separate teammates with explicit file boundaries
- 5
Monitor progress with team-monitor.sh; wait for all teammates to complete before lead merges results
Three parallel reviews completed and consolidated into a single report covering security, performance, and test coverage
Requisitos
- CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS environment variable (set to 1)
- claude CLI installed and authenticated
Instalación
clawhub install matthew-a-gordon/claude-code-teamsEjecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.