team-patterns

name

team-patterns

description

Expert knowledge base of team organization patterns for multi-agent AI systems and software engineering. Provides pattern selection guidance for scenarios including incident response, greenfield design, legacy migration, security audits, performance optimization, API design, data pipelines, testing, documentation, and code migration. This skill should be used when designing agent teams, selecting team topologies, choosing communication patterns, sizing teams, or determining optimal team structure for a given situation. Also use when users mention "team pattern", "team topology", "agent coordination", "team design", "swarm configuration", "multi-agent architecture", "how many agents", or "what agents do I need".

Team Patterns - Expert Team Organization Knowledge

Decision-support index for designing optimal team configurations. Start here, then drill into specific reference files as needed.

Quick Reference: Scenario to Pattern

Scenario	Primary Pattern	Topology	Size	Key Risk
Production incident	Incident Command	Hub-and-spoke	4-6	IC bottleneck
Greenfield design	Squad + ARB	Mesh + periodic star	4-7	Over-engineering
Legacy refactoring	Strangler Fig + enabling team	Pipeline + mesh	5-7	Stalled migration
Security audit	Red/Blue/Purple team	Adversarial + bridge	8-15	Collusion
Performance optimization	Tiger team	Mesh	3-5	Premature scaling
API design	Design-first embedded team	Collaborative mesh	4-6	Approval bottleneck
Data pipeline	Cross-functional DataOps	Mesh + pipeline	4-6	Infra vs. insight
Test suite creation	Integrated quality + pyramid	Mesh (shared)	3-5	QA silo
Documentation overhaul	Docs-as-Code + embedded writers	Mesh + pipeline	3-5	Doc debt
Code migration	Strangler Fig + migration squad	Pipeline + hub-and-spoke	5-8	Dual-system stall

Decision Matrix: Situation Traits to Pattern

When the scenario does not match the quick reference table above, use this trait-based matrix:

Urgency	Complexity	Knowledge	Risk	-> Pattern	-> Topology
Critical	Any	Any	Zero	Incident Command	Hub-and-spoke
High	Complex-entangled	Concentrated	Low	Tiger Team	Mesh
High	Complex-decomposable	Distributed	Low	Fan-Out/Fan-In	Star
Normal	Well-defined	Concentrated	Moderate	Pipeline	Pipeline
Normal	Well-defined	Distributed	Moderate	Supervisor-Worker	Tree
Normal	Complex-decomposable	Distributed	Moderate	Cross-functional Squad	Mesh
Normal	Complex-entangled	Distributed	Low	Mob/Ensemble	Hub+Mesh
Normal	Adversarial	Distributed	Zero	Red/Blue/Purple	Adversarial
Exploratory	Complex-decomposable	Unknown	High	Small Mesh + ARB	Mesh+Star
Exploratory	Complex-entangled	Unknown	Moderate	Debate/Adversarial	Point-to-point+Judge
Any	Any (quality-critical)	Any	Any	Add Reflection Loop	Bidirectional loop

Automated selection: Run scripts/pattern-selector.py --urgency <X> --complexity <Y> for programmatic recommendations.

Pattern Index

#	Pattern	Topology	Key Rule
1	Incident Command	Hub-and-spoke	IC does NOT touch code
2	Tiger Team	Mesh	Full removal from BAU; disband after
3	Architecture Review Board	Star	Embed governance in workflow, not gates
4	Mob/Ensemble	Hub+Mesh	Ideas must pass through another's hands
5	Pair Programming	Point-to-point	Driver-Navigator, Ping-Pong, Strong-Style
6	Red/Blue/Purple	Adversarial	Verifier agent blocks collusion
7	Strangler Fig	Pipeline+Hub	Never big-bang rewrite
8	Fan-Out/Fan-In	Star	Workers must be truly independent
9	Supervisor-Worker	Tree	Max 2 hierarchy levels
10	Debate/Adversarial	P2P+Judge	Voting improves reasoning 13.2%
11	Blackboard	Star+shared	Clear read/write protocols
12	Reflection/Self-Critique	Loop	78.6% -> 97.1% accuracy with reflection

Organizational Theory Constraints

Principle	Rule	Number
Communication overhead	n(n-1)/2 channels	7 agents = 21 channels
Optimal team size	Harvard Business School	4.6 members optimal
Mesh ceiling	Never exceed	7 agents in mesh
Dunbar's working limit	Close relationships	9 max (sub-teams form beyond)
Decision degradation	Bain & Company	-10% effectiveness per member beyond 7
Brooks's Law	Adding to late project	Makes it later
Conway's Law	Team structure shapes architecture	Use Inverse Conway Maneuver
Ringelmann Effect	Agents overlap, not loaf	Clear task boundaries prevent redundancy

Calculator: Run scripts/team-size-calculator.py <N> to assess any team size.

Agent Role Archetypes

Archetype	Purpose	Example Agents
Coordinator	Decomposes work, manages flow	project-manager
Architect	Designs systems, defines interfaces	solid-architect
Implementer	Writes code, builds features	independent-contributor, IC-opus, IC-sonnet
Critic	Finds flaws, validates quality	critical-code-reviewer, complexity-reducer
Investigator	Explores, gathers context	research-methodology-expert
Specialist	Domain expertise	testing-expert, web-bug-hunter, audio-expert
Refactorer	Improves existing code	refactor-master, code-refactor-splitter
Documenter	Creates/maintains documentation	document-maintainer
Releaser	Packages, versions, ships	release-manager

Full mapping: See references/agent-archetype-mapping.md for the complete archetype table — illustrative, for understanding what kinds of roles exist. The team-architect uses these archetypes to describe roles in its design, not to assign specific installed agents (matching to installed agents is the agent-explorer's role, downstream).

Communication Topology Reference

Topology	Channels	Scaling	Max Agents	Primary Use
Hub-and-spoke	n	Linear	~15	Coordination-heavy
Mesh	n(n-1)/2	Quadratic	7	Creative, small teams
Pipeline	n-1	Linear	Unlimited	Sequential processing
Tree	n-1	Log depth	Unlimited	Large teams, hierarchy
Star	2n	Linear	Unlimited	Parallel execution
Adversarial	Fixed	Fixed	By team	Security, validation

Diagrams: See references/topology-diagrams.md for ASCII diagrams of all 6 topologies.

Detailed References

Drill into these files only when you need full research for a specific category:

File	Content	When to Read
references/classic-se-patterns.md	Patterns 1-7: Incident Command, Tiger Team, ARB, Mob, Pair, Spotify, Team Topologies	Designing teams using SE patterns
references/multi-agent-patterns.md	Patterns 8-12: Supervisor-Worker, Pipeline, Fan-Out, Debate, Voting, Reflection, ToT, Blackboard, MoE, CrewAI, LangGraph	Designing AI agent coordination
references/organizational-theory.md	Conway's Law, Brooks's Law, Tuckman, Ringelmann, Communication Overhead	Justifying team size and structure decisions
references/scenario-playbooks.md	Complete playbooks for all 10 scenarios	When the quick reference table matches a scenario
references/topology-diagrams.md	ASCII diagrams for all 6 topologies with formulas	Including diagrams in team design output
references/agent-archetype-mapping.md	Catalog of archetypes with example agents (illustrative)	Understanding the archetype vocabulary used in role design
references/sources.md	97 source URLs, categorized	Attribution, further reading

Example Team Designs

Complete worked examples matching the team-architect output format:

Example	Pattern	Agents	File
Production API 500 errors	Incident Command + Fan-Out	4-5	examples/incident-response-team.md
Greenfield SaaS application	Squad + ARB	5-6	examples/greenfield-fullstack-team.md
Python 2 to 3 monolith migration	Strangler Fig + Enabling	5	examples/legacy-migration-team.md

Scripts

Script	Purpose	Usage
scripts/pattern-selector.py	Recommends patterns from traits or scenarios	python pattern-selector.py --urgency critical
scripts/team-size-calculator.py	Communication overhead calculator	python team-size-calculator.py 5 8 12
scripts/scan-agents.sh	Lists installed agents with descriptions	bash scan-agents.sh