From wicked-garden
Core agentic architecture patterns and anti-patterns for building robust multi-agent systems. Use when: "agent pattern", "multi-agent design", "how should I structure agents", "agentic anti-pattern"
npx claudepluginhub mikeparcewski/wicked-garden --plugin wicked-gardenThis skill uses the workspace's default tool permissions.
Reference guide for proven agentic architecture patterns and common anti-patterns to avoid.
Provides Ktor server patterns for routing DSL, plugins (auth, CORS, serialization), Koin DI, WebSockets, services, and testApplication testing.
Conducts multi-source web research with firecrawl and exa MCPs: searches, scrapes pages, synthesizes cited reports. For deep dives, competitive analysis, tech evaluations, or due diligence.
Provides demand forecasting, safety stock optimization, replenishment planning, and promotional lift estimation for multi-location retailers managing 300-800 SKUs.
Reference guide for proven agentic architecture patterns and common anti-patterns to avoid.
Agents execute in fixed order. Each agent's output feeds the next.
Use when: Linear workflow, clear dependencies, predictable stages.
Example: Research → Analyze → Draft → Review
Tradeoffs: Simple coordination, but bottlenecks if any agent is slow.
Parent agent delegates to specialized child agents. Parent aggregates results.
Use when: Complex tasks decomposable into subtasks, need centralized coordination.
Example: Project manager agent delegates to designer, developer, QA agents.
Tradeoffs: Clear accountability, scales well, but single point of failure at parent.
Peer agents work together on shared problem. No fixed hierarchy.
Use when: Problem benefits from diverse perspectives, no clear lead agent.
Example: Multiple specialized reviewers collaborating on code review.
Tradeoffs: Rich outputs, but requires conflict resolution and consensus mechanisms.
Agents operate independently with minimal coordination.
Use when: Tasks are fully independent, parallel execution needed.
Example: Multiple monitoring agents checking different services.
Tradeoffs: Maximum parallelism, but no shared learning or coordination.
Human approval gates at critical decision points.
Use when: High-stakes decisions, regulatory requirements, learning from human feedback.
Example: Agent drafts report, human approves before sending.
Tradeoffs: Adds latency, but ensures safety and quality.
Reason → Act → Observe cycle. Agent reasons about next action, executes it, observes result, repeats.
Use when: Dynamic environments, need adaptive behavior based on feedback.
Example: Debugging agent tries fixes, observes test results, adapts approach.
Tradeoffs: Flexible and adaptive, but can be inefficient if reasoning is expensive.
Agent creates full plan upfront, then executes all steps.
Use when: Environment is predictable, planning overhead is justified.
Example: Multi-step data pipeline with known transformations.
Tradeoffs: Efficient execution, but brittle if environment changes.
Agent reviews its own outputs and iteratively improves them.
Use when: Quality matters more than speed, self-improvement is valuable.
Example: Writer agent drafts, critiques, and refines output.
Tradeoffs: Higher quality outputs, but increased token usage and latency.
Is task decomposable into subtasks?
├─ YES: Is there a clear lead/coordinator role?
│ ├─ YES: Use Hierarchical
│ └─ NO: Use Collaborative
└─ NO: Does task require adaptation based on results?
├─ YES: Can you plan all steps upfront?
│ ├─ YES: Use Plan-Execute
│ └─ NO: Use ReAct
└─ NO: Do you have multiple independent tasks?
├─ YES: Use Autonomous
└─ NO: Use Sequential
Add Human-in-the-Loop gates for any high-stakes decisions. Add Reflection layer when quality is critical.
Patterns are composable:
Avoid these combinations:
refs/pattern-catalog.md - Detailed pattern descriptions with implementation examplesrefs/anti-patterns-design.md - Design anti-patterns (God Agent, Tight Coupling, Missing Guardrails, Deep Nesting, No Observability)refs/anti-patterns-operational.md - Operational anti-patterns (Sequential Bottleneck, Context Bloat, Redundant Agents, Hardcoded Prompts, Missing Timeouts)