custom-claude-skills

Generate complete Databricks App projects with MLflow agent server for LangGraph, OpenAI Agents SDK, or non-conversational agents. Use when creating new AI agent applications for Databricks Apps, building conversational or stateless API agents, setting up MLflow tracing for agent observability, or deploying agents to Databricks workspace.

Deploy AI agents (LangGraph, OpenAI SDK, or custom frameworks) to Databricks Model Serving using Databricks Asset Bundles (DAB). Use when deploying agents to Model Serving with infrastructure-as-code, multi-environment management (dev/staging/prod), serverless compute, and optional evaluation jobs.

Create LangChain-compatible tools from Databricks MCP servers (managed and external) for use in any AI agent framework. Use when connecting to Databricks managed MCP (Vector Search, Genie, UC Functions, DBSQL), integrating external MCP servers, or creating tools from Unity Catalog functions for LangGraph/LangChain agents.

Build LangGraph multi-agent systems with intelligent supervisor orchestration of specialized worker agents (Genie, RAG, MCP, LLM, custom). Use when creating multi-agent systems, implementing supervisor patterns, routing queries to specialized agents, or building hierarchical/sequential/parallel agent workflows.

Set up Databricks retrieval tools for AI agents using VectorSearchRetrieverTool (unstructured/RAG) and GenieAgent (structured/SQL). Tool configuration only - no agent implementation. Use when setting up Vector Search retrieval tools, creating Genie tools, configuring unstructured/RAG retrieval, setting up structured data tools, adding retrieval tools to agents, or configuring SQL query tools.

Create Databricks Asset Bundles (DAB) from notebooks or Python files with task dependencies. Supports text descriptions, Mermaid diagrams, and workflow diagram images. Generates modular bundles with serverless compute, variables, and deployment scripts. Use when setting up Databricks workflows, creating jobs with task dependencies, converting pipelines into DABs, or deploying across environments (dev/prod).

Generate Databricks notebooks (.py) with local IDE development support using Databricks Connect. This skill should be used when creating new notebooks for AI agent development, ML/AI workflows, or ETL pipelines that need to work both locally (in VS Code/PyCharm) and in Databricks workspace. Automatically includes dual-environment detection and setup code.

Create clean, well-structured Mermaid diagrams for flowcharts, sequence diagrams, class diagrams, ER diagrams, state diagrams, and architecture visuals. ALWAYS generates both .mermaid source files AND image files (PNG/SVG/PDF) using mermaid CLI. Use when the user asks to create diagrams, visualize workflows, model systems, document APIs, show data relationships, or generate architecture documentation. Handles syntax nuances, layout optimization, and common pitfalls across all Mermaid diagram types.

Create comprehensive competitive battle cards for product managers performing competitive analysis. Use when user requests competitive analysis, battle cards, or product comparisons. Accepts product names and attributes/features to compare (via text or uploaded files), performs web research on each product, and generates detailed battle cards in markdown or Excel format showing how the main product compares against competitors.

Transform documents into structured Jira epics with user stories. Use when user needs to create Jira epics from documents, requirements, or proposals, or wants to break down epics into user stories following the 5-section epic template format (Problem Statement, Proposed Solution, Existing Solutions, Deliverables/Impact, Additional Information).

Manage git worktrees with smart automation. Create worktrees with auto-setup (dependencies, env files), list worktrees with merge status, clean up stale worktrees, and switch between worktrees. Use when working on multiple branches simultaneously, doing quick bug fixes, or reviewing PRs without stashing.

Bootstrap new projects by selecting and integrating skills from custom-claude-skills repository with proper structure, uv environment management, and comprehensive documentation. Use when initializing projects, setting up project scaffolding, integrating multiple skills, or generating project documentation (PROJECT_PLAN, REQUIREMENTS, ARCHITECTURE).

Auto-generate comprehensive unit tests for Python codebases with pytest, coverage reports, and uv package management. Use when creating tests, setting up pytest, analyzing test coverage, or testing Databricks notebooks with local Spark. Generates fixtures, parametrized tests, mocking, and coverage reports.

Auto-format Python code with intelligent tool selection based on file type using uv package management. Uses blackbricks for Databricks notebooks (preserves cell structure), black+isort for regular Python, and ruff for linting. Use when formatting code, fixing style issues, or preparing code for CI/CD.

Convert Claude Code skills to work with Codex CLI and Gemini CLI. Use when users want to make their CLAUDE.md or skill files compatible with other AI coding assistants, share skills with a wider audience, or create multi-agent instruction files. Trigger phrases include "make skill work with codex", "convert to gemini", "cross-platform skill", "multi-agent instructions", "share skill with codex/gemini users".