Choosing Swarm Patterns
Pick the right Agent Relay swarm pattern and write correct workflow/YAML configs
Test report
- Verdict
- Tested · Works
- Score
- Tested
- Jul 21, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 86c0e03
Found SKILL.md at .claude/skills/choosing-swarm-patterns/SKILL.md (also a shorter .agents/ copy); the .claude one is complete. Frontmatter parses with name+description; the dir holds only SKILL.md (no scripts/bundled files to break) and the body's "Source of Truth" citations are external @relayflows/core package refs, not install deps. I verified an honesty claim: all 7 flat MCP tool names (send_dm, check_inbox, list_agents, post_message, reply_to_thread, add_agent, remove_agent) appear in packages/cli/src/cli/agent-relay-mcp.ts (HTTP 200). For output I wrote a cold baseline config and a skill-followed config for the same 3-reviewer fan-out task: baseline invented an `agent-relay`/`new Swarm()` class API that would fail on import, while the skill produced correct `@relayflows/core` `workflow().pattern('fan-out')` fluent-builder code with `interactive:false` workers and `{{steps.X.output}}`+dependsOn aggregation wiring. No security smells; docs are notably candid about unimplemented fields (consensusStrategy vote-tally, hierarchical==hub-spoke, reflectOnBarriers).
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 8/10
- Docs & honesty 5/5
What Choosing Swarm Patterns does
A reference skill for coordinating multiple AI agents with Agent Relay's @relayflows/core workflow engine: a decision framework plus accurate YAML and fluent-builder examples for 24 swarm patterns (fan-out, pipeline, hub-spoke, consensus, mesh, handoff, cascade, dag, debate, hierarchical, and 14 role-driven ones). Triggers when the user is choosing an orchestration topology or authoring an Agent Relay workflow. It also documents real API shapes, the flat MCP tool names, and which config fields are declarative-only versus wired into the runner.
How to install Choosing Swarm Patterns
git clone --depth 1 https://github.com/AgentWorkforce/relay.git /tmp/choosing-swarm-patterns-src
mkdir -p ~/.claude/skills
cp -R /tmp/choosing-swarm-patterns-src/.claude/skills/choosing-swarm-patterns ~/.claude/skills/choosing-swarm-patterns
# Pure-knowledge skill: a single SKILL.md, no scripts or bundled deps to load.
# Actually RUNNING a generated workflow (not needed to use the skill's guidance) requires the engine: npm i @relayflows/core
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Choosing Swarm Patterns
-
/choosing-swarm-patternsPick the right Agent Relay swarm pattern and write correct workflow/YAML configs
It also activates on plain-language prompts like these:
-
Which Agent Relay pattern fits a review-then-fix loop -
Choose between fan-out and pipeline for this workflow -
Design a hub-spoke workflow YAML for these agents
Frequently asked questions
- Is the Choosing Swarm Patterns skill free?
- Yes. The skill itself is free from AgentWorkforce/relay. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Choosing Swarm Patterns work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 21, 2026. Verdict: Tested · Works. Found SKILL.md at .claude/skills/choosing-swarm-patterns/SKILL.md (also a shorter .agents/ copy); the .claude one is complete. Frontmatter parses with name+description; the dir holds only SKILL.md (no scripts/bundled files to break) and the body's "Source of Truth" citations are external @relayflows/core package refs, not install deps. I verified an honesty claim: all 7 flat MCP tool names (send_dm, check_inbox, list_agents, post_message, reply_to_thread, add_agent, remove_agent) appear in packages/cli/src/cli/agent-relay-mcp.ts (HTTP 200). For output I wrote a cold baseline config and a skill-followed config for the same 3-reviewer fan-out task: baseline invented an `agent-relay`/`new Swarm()` class API that would fail on import, while the skill produced correct `@relayflows/core` `workflow().pattern('fan-out')` fluent-builder code with `interactive:false` workers and `{{steps.X.output}}`+dependsOn aggregation wiring. No security smells; docs are notably candid about unimplemented fields (consensusStrategy vote-tally, hierarchical==hub-spoke, reflectOnBarriers).
- What is the Choosing Swarm Patterns SkillProof Score?
- 9.2/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 8/10, docs & honesty 5/5.
- How do I install Choosing Swarm Patterns?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Choosing Swarm Patterns with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.