What problem does it solve? Building on Amazon Bedrock involves five separate API endpoints, provider-specific request formats, and frequent model changes that cause errors like ThrottlingException, AccessDeniedException, and malformed requests. This Skill provides authoritative, up-to-date procedures and reference files so agents invoke models, build RAG pipelines, and deploy agents correctly the first time. ## Core Features & Use Cases - Model Invocation: Guides Converse API vs InvokeModel selection, enforces explicit maxTokens to prevent silent quota reservation, and provides Python/TypeScript SDK patterns. - Knowledge Bases (RAG): Step-by-step KB creation, chunking strategy selection, and retrieval modes (retrieve-and-generate vs retrieve vs manual). - Agents & AgentCore: Procedures for creating agents with action groups, deploying to AgentCore Runtime, exposing APIs via Gateway, and setting up AgentCore Payments (x402 microtransactions with Coinbase CDP or Stripe Privy). - Operations & Troubleshooting: Prompt caching setup and debugging, quota health audits, cost tracking by team/model, Claude generation migration (4.5 to 4.6 to 4.7), and a decision table for common errors. - Use Case: A developer hits repeated ThrottlingException errors; the Skill diagnoses unset maxTokens reserving excess quota, checks service quotas, and recommends cross-region inference profiles. ## Quick Start Ask the agent to help you invoke a Claude model on Bedrock, set up a Knowledge Base for your documents, or diagnose a specific Bedrock error you are seeing.
