What problem does it solve? Manually assembling project documentation (READMEs, architecture docs, runbooks, KT docs) at the end of a data engineering pipeline is slow and error-prone. This Skill acts as Agent 13 of 15 in the Databricks Multi-Agent SDLC Platform, automatically synthesizing all documentation from the artifacts produced by the twelve upstream agents. ## Core Features & Use Cases - Artifact-grounded documentation: Reads prior agent artifacts from /Volumes/{catalog}/state/artifacts/{ticket}/ and summarizes them without re-deriving facts, ensuring docs match what was actually built. - Contract-validated output: Returns JSON conforming to a strict output contract (status, artifact_path, citations, tokens) that the LangGraph orchestrator validates before handoff to the Deployment Agent. - Databricks backend integration: Calls the Foundation Model API (e.g., databricks-claude-sonnet-4-5) via Model Serving for generation, with guidance on why Genie Code cannot be invoked headlessly. - Use Case: After agents 1-12 complete requirement analysis, data modeling, PySpark/SQL development, and testing for a Jira ticket, this agent produces the README, technical spec, deployment guide, runbook, user guide, and KT document, writes them to the lakehouse, and logs the run to state.agent_runs. ## Quick Start Ask the agent to generate the full documentation set for ticket DATA-1234 from the upstream artifacts in the Databricks volume.
