databricks-core-workflow-b

Execute Databricks secondary workflow: MLflow model training and deployment. Use when building ML pipelines, training models, or deploying to production. Trigger with phrases like "databricks ML", "mlflow training", "databricks model", "feature store", "model registry".

Quality

—

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Passed

No known issues

Quality

Content

80%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content is a lean, fully executable ML workflow with strong actionability and minimal padding, but it is undermined by missing validation checkpoints on a destructive batch write and by an unreferenced, duplicated bundle file. Workflow clarity and progressive disclosure are the weakest dimensions for those reasons.

Suggestions

Add a validation/checkpoint step before the Step 6 batch overwrite (e.g. verify champion model loads and preview row counts, then confirm before `mode("overwrite")`) to satisfy the destructive/batch feedback-loop requirement.

Either link references/implementation-guide.md from the body (e.g. 'For full implementation details, see [implementation-guide.md](references/implementation-guide.md)') and keep the body as an overview, or remove the redundant guide to avoid duplication.

Add a brief verify-then-proceed note to model promotion (Step 3) so the champion alias is confirmed before serving deployment.

Dimension	Reasoning	Score
Conciseness	The body is lean and code-driven: each section is a focused, executable step with minimal prose and no explanation of concepts Claude already knows (no 'what is MLflow' padding). Despite covering 6 steps it stays efficient and assumes competence. Not 2 because there is no unnecessary explanation to trim.	3 / 3
Actionability	Fully executable, copy-paste-ready Python throughout, using real SDK calls with concrete identifiers, plus a hyperparameter sweep example and an error-handling table mapping errors to specific solutions. Not 2 because the code is complete rather than pseudocode with missing details.	3 / 3
Workflow Clarity	Six steps are clearly sequenced, but Step 6 is a batch operation that overwrites a Delta table (`mode("overwrite")`, `saveAsTable`) with no validation checkpoint or feedback loop, and no step includes validate-then-proceed gating. The rubric caps workflow clarity at 2 when validation is missing for destructive/batch operations. Not 3 because of those missing checkpoints; not 1 because the sequence itself is clear.	2 / 3
Progressive Disclosure	A references/implementation-guide.md bundle exists but is never linked or signaled in the body, and the body keeps all detail inline in a monolithic walkthrough (the guide largely duplicates Step 1). References are present but not clearly signaled, and content that could be split out stays inline. Not 1 because the body is section-organized; not 3 because the reference file is not surfaced and inline detail could be offloaded.	2 / 3
	Total	10 / 12 Passed

Description

82%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description clearly states both capability and trigger conditions with a solid set of natural trigger phrases, making it highly complete and trigger-aware. It is somewhat held back by high-level rather than granular action specificity and by broad terms that risk overlapping with sibling Databricks skills.

Suggestions

Replace the generic 'databricks ML' trigger with more distinctive phrases (e.g. 'mlflow model registry', 'champion challenger alias', 'databricks model serving endpoint') to reduce overlap with databricks-core-workflow-a.

Expand the 'what' clause to enumerate concrete actions (e.g. 'register and alias model versions, deploy serving endpoints, run batch inference jobs') to raise specificity.

Tighten the trigger list to the terms a user would most naturally say when only this skill applies.

Dimension	Reasoning	Score
Specificity	Names the domain and a couple of actions ('MLflow model training and deployment', 'building ML pipelines, training models, or deploying to production'), but these are high-level rather than a list of multiple concrete distinct actions. Not 1 because concrete actions are named; not 3 because it stops at training/deployment breadth rather than enumerating specific operations.	2 / 3
Completeness	Explicitly answers both 'what' ('MLflow model training and deployment') and 'when' ('Use when building ML pipelines, training models, or deploying to production') with explicit trigger guidance. Both halves are clearly present.	3 / 3
Trigger Term Quality	Provides an explicit set of natural trigger phrases users would say ('databricks ML', 'mlflow training', 'databricks model', 'feature store', 'model registry'), giving good coverage across the ML/Databricks domain. These are natural terms rather than technical jargon.	3 / 3
Distinctiveness Conflict Risk	Triggers are tied to MLflow/feature-store/registry, which is somewhat specific, but the deprecated banner itself flags 'no specific pain cluster' and broad terms like 'databricks ML' could overlap with sibling Databricks skills (e.g. databricks-core-workflow-a). Not 1 because it has a clear ML niche; not 3 because overlap with adjacent Databricks skills remains.	2 / 3
	Total	10 / 12 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: jeremylongshore/claude-code-plugins-plus-skills
Commit: 02d6341

Reviewed: 4 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.