Skills that extend Databricks Assistant (or other coding agents like Claude Code) with Databricks-specific features or workflows.
| Skill | Slash Command | Description |
|---|---|---|
| genie-space-diagnostics | /genie-space-diagnostics |
Audit Genie Space configuration against best practices with prioritized remediation plan |
| genie-space-optimizer | /genie-space-optimizer |
Benchmark-driven iterative optimization of Genie Space accuracy using the Benchmark API |
| multi-agent-architecture | /multi-agent-architecture |
Design multi-agent architectures for PoC/hackathon/MVP projects with Excalidraw diagrams |
| parse-documents | — | Parse and chunk documents for ingestion |
Use manage-skills.sh to symlink skills into any project's .claude/skills/ directory. Skills are auto-triggered by Claude when relevant to the conversation.
git clone https://github.com/hiydavid/databricks-agent-skills.git
cd databricks-agent-skills
# List available skills
./scripts/manage-skills.sh list
# Add all skills to a project
./scripts/manage-skills.sh add --all --to ~/my-project
# Add a single skill
./scripts/manage-skills.sh add multi-agent-architecture --to ~/my-project
# Check what's installed
./scripts/manage-skills.sh status --in ~/my-project
# Remove a skill
./scripts/manage-skills.sh remove multi-agent-architecture --from ~/my-projectIf --to/--from/--in is omitted, it defaults to the current working directory.
Copy the desired skill folder into your workspace skills directory:
/Users/{username}/.assistant/skills/
└── {skill-name}/
├── SKILL.md
├── scripts/
└── references/
The Assistant automatically discovers skills in agent mode. See the Databricks docs for details.
databricks-agent-skills/
├── genie-space-diagnostics/ # Static audit of Genie Space configuration
│ ├── SKILL.md
│ ├── scripts/
│ └── references/
├── genie-space-optimizer/ # Benchmark-driven iterative optimization
│ ├── SKILL.md
│ ├── scripts/
│ └── references/
├── multi-agent-architecture/
├── parse-documents/
├── scripts/ # Dev tooling
│ ├── manage-skills.sh
│ └── mermaid-to-excalidraw.mjs
├── test-skills/ # Test scenarios for validating skills
│ └── equity-analysis.md
└── README.md
- Create the skill folder:
mkdir -p my-skill - Add
my-skill/SKILL.mdwith instructions and references - Add a slash command: create
commands/my-skill.mdwith frontmatter (description,argument-hint) that loads the SKILL.md - Install and test:
./scripts/manage-skills.sh add my-skill --to ~/my-project
- Add skills to a test project:
./scripts/manage-skills.sh add --all --to ~/my-project - Start a new Claude Code session in that project
- Invoke via slash command (e.g.,
/multi-agent-architecture) or trigger naturally - Edit the skill files, start a new session, and re-test (symlinks pick up changes automatically)
- Clone the repo and create a branch
- Follow the "Adding a new skill" steps above
- Test locally, then open a PR