| name | project-starter |
| description | 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). |
Project Starter Skill
Bootstrap new projects by selecting and integrating skills from the custom-claude-skills repository, setting up proper project structure, and generating comprehensive documentation.
Repository Reference
Skill Catalog
Databricks Platform (4)
- databricks-asset-bundle - DAB configs with serverless compute and multiple input formats
- databricks-local-notebook - Notebooks with local IDE development via Databricks Connect
- databricks-agent-deploy2app - Deploy AI agents to Databricks Apps
- databricks-agent-deploy-model-serving-dab - Deploy agents to Model Serving using DABs
LangGraph (4)
- langgraph-genie-agent - Agents with Databricks Genie API
- langgraph-unstructured-tool-agent - RAG agents with Vector Search
- langgraph-multi-agent-supervisor - Multi-agent systems with supervisor orchestration
- langgraph-mcp-tool-calling-agent - Agents with MCP tools
Python (2)
- pytest-test-creator - Auto-generate unit tests with coverage
- python-code-formatter - Format code (blackbricks for Databricks, black+isort for regular)
Planning & Visualization (1)
- mermaid-diagrams-creator - Mermaid diagrams with automatic PNG/SVG/PDF generation
General (2)
- jira-epic-creator - Transform documents into structured Jira epics
- battle-card-creator - Competitive battle card creation
Project Structure
project-name/
├── .claude/
│ ├── skills-repo/ # Git submodule of custom-claude-skills
│ ├── skills/ # Symlinks to selected skills
│ └── project-context.md # Generated context for this project
├── docs/
│ ├── PROJECT_PLAN.md
│ ├── REQUIREMENTS.md
│ └── ARCHITECTURE.md # If applicable
├── src/
├── tests/
├── notebooks/ # If using Databricks skills
├── configs/
├── .python-version
├── pyproject.toml # uv managed
├── uv.lock
├── .gitignore
└── README.md
Initialization Workflow
Step 1: Understand Requirements
Gather: project name, primary goal, skills needed, tech stack preferences.
Step 2: Select Skills
Recommend skill combinations based on project type:
| Project Type | Recommended Skills |
|---|
| RAG/Agent | langgraph-unstructured-tool-agent, databricks-agent-deploy2app, pytest-test-creator, mermaid-diagrams-creator |
| Multi-Agent | langgraph-genie-agent, langgraph-multi-agent-supervisor, databricks-agent-deploy-model-serving-dab, mermaid-diagrams-creator |
| Databricks Workflows | mermaid-diagrams-creator, databricks-asset-bundle, databricks-local-notebook, python-code-formatter |
| Documentation/Analysis | mermaid-diagrams-creator, battle-card-creator, jira-epic-creator |
Step 3: Create Project Structure
- Create directories per structure above
- Initialize git repository
- Initialize Python project:
uv init --name project-name --python 3.11
- Add skills repository as submodule:
git submodule add https://github.com/qian-yu-db/custom-claude-skills .claude/skills-repo
git submodule update --init --recursive
- Create symlinks for selected skills:
ln -s ../.claude/skills-repo/[skill-path] .claude/skills/[skill-name]
Step 4: Generate Project Context
Create .claude/project-context.md with: project overview, selected skills and why, integration points, development workflow.
Step 5: Generate Documentation
Generate docs/PROJECT_PLAN.md and docs/REQUIREMENTS.md from templates. See templates/ for the template files.
Step 6: Create Initial Scaffolding
LangGraph projects: src/agent.py, src/tools.py, src/config.py, tests/test_agent.py
uv add langchain langgraph langchain-community
Databricks projects: notebooks/01_setup.py, configs/databricks.yml, src/utils.py
uv add databricks-sdk databricks-connect
General projects: src/main.py, tests/test_main.py, configs/config.yaml
uv add --dev pytest pytest-cov ruff
Step 7: Generate README.md
Include: quick start, prerequisites, installation with uv sync, project structure, skills used, links to docs.
Step 8 (Optional): Cross-Platform Support
If user requests --cross-platform, generate instruction files for Claude Code, Codex CLI, and Gemini CLI from a canonical instructions.md. Uses cross-platform-skill/scripts/convert_skill.py for conversion.
| Platform | Instruction File | Commands Dir |
|---|
| Claude Code | CLAUDE.md | .claude/commands/ |
| Codex CLI | AGENTS.md | .codex/prompts/ |
| Gemini CLI | GEMINI.md | .gemini/commands/ |
Commands
Initialize New Project
"Create a new project called [name] for [purpose]" → Execute Steps 1-7.
Add Skills to Existing Project
"Add [skill-name] to this project" → Add submodule if needed, create symlink, update context and requirements.
Generate Project Documentation
"Generate/update project documentation" → Read skill files, analyze code, generate/update docs.
Validate Project Setup
"Validate project setup" → Verify directory structure, submodule status, symlinks, dependencies.
References