From teampowers
Architectural overview of how agents communicate — each agent's specific paths are in its own skill file
How this skill is triggered — by the user, by Claude, or both
Slash command
/teampowers:communication-meshThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
- **Name**: communication-mesh
Teampowers agents communicate in a mesh network. Agents talk directly to each other rather than routing everything through the lead. The lead coordinates but doesn't bottleneck.
This is a reference document. Each agent's specific communication paths (who they receive from, who they send to, and message formats) are defined in that agent's own skill file. You do NOT need to load this document to know how to communicate — check your own skill file.
Only the lead spawns teammates — this is an agent teams constraint. Other agents request new teammates by messaging the lead:
All verification happens in the dev's worktree before merge:
Lead assigns → Scout explores → Scout sends context to Dev
Dev implements (in worktree)
→ Tester validates (in dev's worktree)
→ Reviewer reviews (in dev's worktree)
→ CI runs pipeline (in dev's worktree)
→ CI signals Lead: ready to merge
Lead merges worktree branch → CI runs integration check on merged branch
┌──────────┐
│ Lead │
│(coord.) │
└────┬─────┘
│ assigns tasks, merges worktrees
┌────────────┼────────────┐
▼ ▼ ▼
┌─────────┐ ┌─────────┐ ┌──────────┐
│ Scout 1 │ │ Scout 2 │ │ Ad Hoc │
└────┬────┘ └────┬────┘ └────┬─────┘
│ │ │
▼ ▼ ▼
┌─────────┐ ┌─────────┐ │ domain guidance
│ Dev 1 │ │ Dev 2 │◄─────┘
│(wt-1) │ │(wt-2) │
└────┬────┘ └────┬────┘
│ │ all verification in
▼ ▼ dev's worktree
┌────────┐ ┌────────┐
│ Tester │ │ Tester │ (same tester, different worktrees)
└───┬────┘ └───┬────┘
▼ ▼
┌────────┐ ┌────────┐
│Reviewer│ │Reviewer│ (same reviewer, different worktrees)
└───┬────┘ └───┬────┘
▼ ▼
┌────────┐ ┌────────┐
│ CI │ │ CI │ (same CI, different worktrees)
└───┬────┘ └───┬────┘
│ │
▼ ▼
ready to merge → Lead
Each agent's skill file contains its own:
See:
teampowers:dev — Dev agent communicationteampowers:scout — Scout agent communicationteampowers:tester — Tester agent communicationteampowers:reviewer — Reviewer agent communicationteampowers:ci — CI agent communicationteampowers:ad-hoc-agents — Ad hoc agent communicationnpx claudepluginhub johnhanks1/vim-setup --plugin teampowersStructured messaging protocols for agent team communication: message type selection, plan approval, shutdown procedures, and anti-patterns to avoid.
Provides patterns for multi-agent systems in Claude Code: job description method, shared folder communication, handbook consolidation, context management. Use for complex agent orchestrations.
Use when dispatching subagents, composing prompts for teammates, structuring handoff reports, or managing context boundaries between agents. Covers both subagent prompts and team-level messaging.