Add atlan-lakehouse AI agent skill

[benhuds]

Summary

• New AI agent skill (skills/atlan-lakehouse/) that teaches coding agents how to connect to and query the Atlan Lakehouse
• Cross-platform: Snowflake (Cortex Code), Databricks (Genie Code), and Python (PyIceberg) with directive routing rules that enforce native SQL when a SQL environment is available
• ~45 SQL templates: Entity metadata (completeness, lineage, glossary export) and usage analytics (active users, feature adoption, engagement depth, retention, customer health scoring)
• Storage-agnostic: Installs pyiceberg[s3fs,adlfs,gcsfs] to support S3, Azure, and GCS backends; Polaris vends temporary storage credentials automatically
• Platform intricacies handled: case sensitivity (Polaris lowercase vs Snowflake/Databricks uppercase), namespace fallback (entity_metadata → atlan-ns for pre-Feb 2026 tenants), customer-defined database/catalog names
• Updated repo README with AI Agent Skill section and skills/ directory in repo structure

Test plan

• Install skill in Claude Code (~/.claude/skills/atlan-lakehouse/) and verify it activates on lakehouse-related prompts
• Test from Snowflake/Cortex Code — verify agent uses native SQL, does not invoke PyIceberg
• Test from plain Claude Code terminal — verify agent uses PyIceberg path and asks for credentials
• Spot-check SQL templates against a real lakehouse tenant (health scorecard, retention cohort, glossary export)
• Verify agent handles entity_metadata → atlan-ns fallback correctly

🤖 Generated with Claude Code