LLM App Patterns Skill

When to Activate

• Building multi-step LLM pipelines (chaining calls, conditional routing)
• Adding AI features to an existing application
• Reliability or cost concerns (retries, fallbacks, caching)
• Implementing streaming responses to the client
• Adding guardrails (input validation, output filtering, content policy)
• Designing human-in-the-loop review flows
• Debugging latency spikes or unexpected LLM errors in production

Orchestration Patterns

Sequential chain

Each step's output becomes the next step's input.