# Orchestrate Multi-Agent Claude Flow Swarms

Large development tasks are hard to split, assign, and monitor across agents. This skill provides a phased Claude Flow process for planning, execution, monitoring, and synthesis.

## Install

```bash
npx skillstore add dnyoussef/when-orchestrating-swarm-use-swarm-orchestration
```

## Metadata

- Status: approved
- Slug: dnyoussef-when-orchestrating-swarm-use-swarm-orchestration
- Version: 1.0.0
- Author version: 1.0.0
- Skillstore revision: r1
- Version status: valid
- Tree hash: 0de36e85b09f276bd58b8a71985d5c68289e9bd5481d10cd300cfc4284806a41
- Author: DNYoussef
- GitHub username: DNYoussef
- License: MIT
- Repository: https://github.com/DNYoussef/ai-chrome-extension/tree/main/.claude/skills/workflow/when-orchestrating-swarm-use-swarm-orchestration
- Ref: c579efd02184f1128d6741d16e6c8d951d7d4208
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/dnyoussef-when-orchestrating-swarm-use-swarm-orchestration
- Manifest: https://skillstore.pages.dev/api/skills/dnyoussef-when-orchestrating-swarm-use-swarm-orchestration/manifest

## Capabilities

- Breaks complex work into phased task trees with dependency checks.
- Selects a swarm topology based on planned task count.
- Defines coordinator and specialist agent roles for Claude Flow.
- Assigns tasks after dependency status checks.
- Tracks progress, blocked tasks, agent health, and workload metrics.
- Collects, groups, validates, reports, and archives orchestration results.

## Use Cases

- Coordinate a large application build: Plan architecture, backend, frontend, infrastructure, testing, and review work across specialized agents.
- Standardize agent workflow execution: Use the phase checklist to keep decomposition, assignment, monitoring, and synthesis consistent across projects.
- Review swarm execution quality: Collect progress metrics, blocked tasks, final outputs, and performance targets after a multi-agent run.

## Prompt Templates

### Decompose a Task

```
Use this skill to decompose this task into phases, dependencies, and suggested agent roles: [task].
```

### Prepare a Swarm Plan

```
Create a Claude Flow swarm plan for [project]. Include topology, coordinators, specialist agents, dependencies, and validation checks.
```

### Monitor Active Execution

```
Use this skill to define monitoring steps for the active swarm. Focus on completion rate, blocked tasks, agent health, and progress reporting.
```

### Recover and Synthesize Results

```
Analyze this swarm state: [status]. Recommend recovery actions, final result collection, synthesis groups, validation checks, and report contents.
```

## Limitations

- Requires Claude Flow to be installed and configured separately.
- Includes shell command examples that need review before execution.
- Does not verify task security, package trust, or project policy compliance.
- Can create many agents and long-running monitors if limits are not adjusted.

## Best Practices

- Review every shell command before running it in a project workspace.
- Set explicit agent, duration, and output limits before orchestration starts.
- Validate dependencies and final outputs before accepting synthesized results.

## Anti Patterns

- Do not run unpinned package commands in sensitive repositories without approval.
- Do not spawn large swarms before defining task scope and success criteria.
- Do not treat synthesized output as correct without human review.

## Security Audit

- Audited at: 2026-07-05T13:19:42.894\+00:00
- Summary: The skill is not malicious, and no prompt injection text was found. It does contain real external command risk because many examples run npx claude-flow@alpha, spawn agents, monitor tasks, and write artifacts. Several static hits were markdown, JSON, JavaScript, shebang, jq, date, or workflow-status false positives.

## Stats

- Views: 279
- Downloads: 6
- Favorites: 0
- Popularity score: 0