feat: add amazon-bedrock guided MCP power

[dgallitelli]

Summary

• Add amazon-bedrock Guided MCP Power for Amazon Bedrock
• 7 steering files covering model catalog, Converse API, agents, knowledge bases, guardrails, model customization, and prompts/flows
• Integrates awslabs.bedrock-kb-retrieval-mcp-server and awslabs.aws-bedrock-custom-model-import-mcp-server
• Model catalog includes all foundation model IDs and CRIS inference profiles pulled from live AWS APIs (2025-03-25)
• Always-on guardrails table in POWER.md prevents common mistakes (wrong client, wrong API, missing CRIS)

Closes #82

Steering files

File	Topic
model-catalog.md	All model IDs, CRIS profiles by provider, embedding models
converse-api.md	Unified invocation, streaming, tool use, multimodal, caching
agents.md	Action groups, Lambda, return control, code interpreter
knowledge-bases.md	Vector stores, data sources, chunking, RAG queries
guardrails.md	Content filters, PII, topic denial, grounding
model-customization.md	Fine-tuning, reinforcement, distillation, import
prompts-and-flows.md	Versioned prompts, variables, multi-step flows

Design decisions

• Guided MCP Power with two MCP servers (KB retrieval + custom model import)
• Excludes AgentCore MCP — already covered by the aws-agentcore power
• Excludes Data Automation MCP — deprecated by awslabs
• CRIS-first approach — all examples default to us.* inference profile IDs
• Model catalog from live APIs — not hardcoded documentation, reflects actual availability

Test plan

• Install as local custom power in Kiro
• Verify steering files load correctly via readSteering action
• Test Converse API code generation uses CRIS model IDs
• Test KB retrieval MCP tools connect and query
• Test agent creation workflow generates correct boto3 code

🤖 Generated with Claude Code