session-summarize

What problem does it solve?

When working with a data engineering agent across multiple chat sessions, valuable insights like validated SQL queries, business rules, metric definitions, and user preferences are often lost when a session ends, leading to repeated work and inconsistent context for future interactions.

Core Features & Use Cases

• Session Takeaway Harvesting: Automatically scans the current chat transcript to collect all high-value content including business facts/rules, accepted SQL queries, refined metric definitions, durable user preferences, and reusable multi-step workflows.
• Intelligent Storage Classification: Uses the storage-classify skill's decision tree to route each harvested candidate to the correct store (semantic models, metrics, reference SQL, knowledge base, memory, or AGENTS.md) with the appropriate generation mechanism.
• Confirmation-First Persistence: Presents a deduplicated Summary Manifest of all planned persistence actions for user approval before any heavy generation or file writes occur, avoiding unnecessary work and unintended changes.
• Use Case: After a session where you validated a regional GMV ranking SQL query and defined a custom order status encoding rule, use this skill to capture both the SQL example, the metric definition, and the business rule for future reuse across all your data workflows.

Quick Start

Invoke the session-summarize skill at the end of your working session to harvest, classify, and route all valuable takeaways for persistence after your confirmation.