choosing-swarm-patterns

Name: choosing-swarm-patterns
Author: AgentWorkforce

769stars58forks

How to install choosing-swarm-patterns?

agentworkforce/relay/choosing-swarm-patterns

$npx -y skills add AgentWorkforce/relay --skill choosing-swarm-patterns

Installs into the current project.

Summary

Use when coordinating multiple AI agents with Agent Relay's workflow engine and need to pick the right orchestration pattern - covers the 10 core patterns (fan-out, pipeline, hub-spoke, consensus, mesh, handoff, cascade, dag, debate, hierarchical) plus 14 specialized ones, with d

Files · 1

View on GitHub

SKILL.md

1	### Overview
2
3	The Agent Relay workflow engine (`@relayflows/core`) supports 24 swarm patterns via a single `swarm.pattern` field. Patterns are configured declaratively in YAML or programmatically via the `workflow()` fluent builder — there are no standalone `fanOut(...)` / `hubAndSpoke(...)` helpers. Pick the simplest pattern that solves the problem; add complexity only when the system proves it's insufficient.
4
5	### Two ways to run a pattern
6
7	#### 1. YAML (portable):
8
9	```ts
10	import { runWorkflow } from '@relayflows/core';
11
12	const run = await runWorkflow('workflows/feature-dev.yaml', {
13	vars: { task: 'Add OAuth login' },
14	});
15	```
16
17	#### 2. Fluent builder (programmatic):
18
19	```ts
20	import { workflow } from '@relayflows/core';
21
22	const run = await workflow('feature-dev')
23	.pattern('hub-spoke')
24	.channel('swarm-feature-dev')
25	.agent('lead', { cli: 'claude', role: 'lead' })
26	.agent('developer', { cli: 'codex', role: 'worker', interactive: false })
27	.step('plan', { agent: 'lead', task: 'Plan {{task}}' })
28	.step('implement', { agent: 'developer', task: 'Implement: {{steps.plan.output}}', dependsOn: ['plan'] })
29	.run();
30	```
31
32	Both paths hit the same `WorkflowRunner`.
33
34	### Quick Decision Framework
35
36	```
37	Is the task independent per agent?
38	YES → fan-out (parallel workers, hub collects)
39
40	Does each step need the previous step's output?
41	YES → Is it strictly linear?
42	YES → pipeline
43	NO → dag (parallel where possible, `dependsOn` edges)
44
45	Does a coordinator need to stay alive and adapt?
46	YES → hub-spoke (single-level hub + workers)
47	hierarchical (structurally identical in current impl; use for naming/intent)
48
49	Is the task about making a decision?
50	YES → Do agents need to argue opposing sides?
51	YES → debate (adversarial, full mesh)
52	NO → consensus (cooperative, full mesh + coordination.consensusStrategy)
53
54	Does the right specialist emerge during processing?
55	YES → handoff (sequential chain, one active at a time)
56
57	Do all agents need to freely collaborate?
58	YES → mesh (full peer-to-peer edges)
59
60	Is cost the primary concern?
61	YES → cascade (chain of increasingly capable agents; each step's prompt
62	decides whether to pass through or redo the prior output)
63	```
64
65	### Pattern Reference (Core 10)
66
67	\| # \| Pattern \| Topology (actual edges) \| Best For \|
68	\| --- \| ---------------- \| ------------------------------------------------------------------------------------------------------------------- \| --------------------------------------------------------------------------- \|
69	\| 1 \| fan-out \| Hub broadcasts to N workers; workers reply to hub only \| Independent subtasks (reviews, research, tests) \|
70	\| 2 \| pipeline \| Linear chain (agenti → agent{i+1}) \| Ordered stages (design → implement → test) \|
71	\| 3 \| hub-spoke \| Hub ↔ spokes (bidirectional); no spoke-to-spoke \| Dynamic coordination, lead reviews/adjusts \|
72	\| 4 \| consensus \| Full mesh; decision via `coordination.consensusStrategy` \| Architecture decisions, approval gates \|
73	\| 5 \| mesh \| Full mesh (every agent ↔ every other) \| Brainstorming, collaborative debugging \|
74	\| 6 \| handoff \| Chain; passes control forward \| Triage, specialist routing \|
75	\| 7 \| cascade \| Chain of `dependsOn` steps; all run on success, downstream skipped on upstream failure (no built-in "fall through") \| Cost optimization: cheap first, each step's prompt passes through or redoes \|
76	\| 8 \| dag \| Edges from step `dependsOn` \| Mixed dependencies, parallel where possible \|
77	\| 9 \| debate \| Full mesh (same topology as mesh; roles drive behavior)

People who liked this also viewed

6 picks

Type

choosing-swarm-patterns

769stars58forks

View on GitHub

How to install choosing-swarm-patterns?

agentworkforce/relay/choosing-swarm-patterns

$npx -y skills add AgentWorkforce/relay --skill choosing-swarm-patterns

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	### Overview
2
3	The Agent Relay workflow engine (`@relayflows/core`) supports 24 swarm patterns via a single `swarm.pattern` field. Patterns are configured declaratively in YAML or programmatically via the `workflow()` fluent builder — there are no standalone `fanOut(...)` / `hubAndSpoke(...)` helpers. Pick the simplest pattern that solves the problem; add complexity only when the system proves it's insufficient.
4
5	### Two ways to run a pattern
6
7	#### 1. YAML (portable):
8
9	```ts
10	import { runWorkflow } from '@relayflows/core';
11
12	const run = await runWorkflow('workflows/feature-dev.yaml', {
13	vars: { task: 'Add OAuth login' },
14	});
15	```
16
17	#### 2. Fluent builder (programmatic):
18
19	```ts
20	import { workflow } from '@relayflows/core';
21
22	const run = await workflow('feature-dev')
23	.pattern('hub-spoke')
24	.channel('swarm-feature-dev')
25	.agent('lead', { cli: 'claude', role: 'lead' })
26	.agent('developer', { cli: 'codex', role: 'worker', interactive: false })
27	.step('plan', { agent: 'lead', task: 'Plan {{task}}' })
28	.step('implement', { agent: 'developer', task: 'Implement: {{steps.plan.output}}', dependsOn: ['plan'] })
29	.run();
30	```
31
32	Both paths hit the same `WorkflowRunner`.
33
34	### Quick Decision Framework
35
36	```
37	Is the task independent per agent?
38	YES → fan-out (parallel workers, hub collects)
39
40	Does each step need the previous step's output?
41	YES → Is it strictly linear?
42	YES → pipeline
43	NO → dag (parallel where possible, `dependsOn` edges)
44
45	Does a coordinator need to stay alive and adapt?
46	YES → hub-spoke (single-level hub + workers)
47	hierarchical (structurally identical in current impl; use for naming/intent)
48
49	Is the task about making a decision?
50	YES → Do agents need to argue opposing sides?
51	YES → debate (adversarial, full mesh)
52	NO → consensus (cooperative, full mesh + coordination.consensusStrategy)
53
54	Does the right specialist emerge during processing?
55	YES → handoff (sequential chain, one active at a time)
56
57	Do all agents need to freely collaborate?
58	YES → mesh (full peer-to-peer edges)
59
60	Is cost the primary concern?
61	YES → cascade (chain of increasingly capable agents; each step's prompt
62	decides whether to pass through or redo the prior output)
63	```
64
65	### Pattern Reference (Core 10)
66
67	\| # \| Pattern \| Topology (actual edges) \| Best For \|
68	\| --- \| ---------------- \| ------------------------------------------------------------------------------------------------------------------- \| --------------------------------------------------------------------------- \|
69	\| 1 \| fan-out \| Hub broadcasts to N workers; workers reply to hub only \| Independent subtasks (reviews, research, tests) \|
70	\| 2 \| pipeline \| Linear chain (agenti → agent{i+1}) \| Ordered stages (design → implement → test) \|
71	\| 3 \| hub-spoke \| Hub ↔ spokes (bidirectional); no spoke-to-spoke \| Dynamic coordination, lead reviews/adjusts \|
72	\| 4 \| consensus \| Full mesh; decision via `coordination.consensusStrategy` \| Architecture decisions, approval gates \|
73	\| 5 \| mesh \| Full mesh (every agent ↔ every other) \| Brainstorming, collaborative debugging \|
74	\| 6 \| handoff \| Chain; passes control forward \| Triage, specialist routing \|
75	\| 7 \| cascade \| Chain of `dependsOn` steps; all run on success, downstream skipped on upstream failure (no built-in "fall through") \| Cost optimization: cheap first, each step's prompt passes through or redoes \|
76	\| 8 \| dag \| Edges from step `dependsOn` \| Mixed dependencies, parallel where possible \|
77	\| 9 \| debate \| Full mesh (same topology as mesh; roles drive behavior)

People who liked this also viewed

6 picks

Type