Content
64%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a solid, actionable skill with concrete YAML configurations and executable commands that cover the full lifecycle of Lakeflow Jobs development. Its main weaknesses are the lack of error recovery/validation feedback loops in the workflow, and some content bloat from inline templates and basic Spark code that Claude already knows. The structure would benefit from tighter editing and explicit checkpoints in the deployment workflow.
Suggestions
Add explicit error recovery guidance to the Development Workflow section (e.g., 'If validate fails, check for missing variable references in databricks.yml; if deploy fails, verify profile permissions with `databricks auth describe`').
Remove or significantly trim the notebook code section — Claude already knows basic Spark read/write/SQL patterns; instead, focus only on Databricks-specific patterns like dbutils.widgets.get().
Consider extracting the CLAUDE.md/AGENTS.md template content into a separate referenced file to reduce the main skill's token footprint.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The content is mostly efficient with good YAML/code examples, but includes some unnecessary sections like the full CLAUDE.md/AGENTS.md template content (which is boilerplate), and the project structure diagram is somewhat redundant given the scaffolding command generates it. The notebook code section explains basic Spark operations Claude already knows. | 2 / 3 |
Actionability | Provides fully executable commands (bundle init with config-file), complete YAML configurations for job definitions, scheduling, and multi-task dependencies, and concrete Python code for notebook tasks. The scaffolding command is copy-paste ready with clear parameter explanations. | 3 / 3 |
Workflow Clarity | The Development Workflow section lists validate → deploy → run → check steps clearly, but lacks explicit validation checkpoints and error recovery feedback loops. There's no guidance on what to do if validation fails, if deployment errors occur, or how to verify job output correctness before promoting. | 2 / 3 |
Progressive Disclosure | References the parent 'databricks-core' skill appropriately and includes external documentation links, but the content is somewhat monolithic — sections like scheduling options, multi-task dependencies, and notebook code patterns could be split into separate reference files. The inline CLAUDE.md template adds bulk that could be referenced externally. | 2 / 3 |
Total | 9 / 12 Passed |