Content
72%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-structured router-style SKILL.md that excels at progressive disclosure, with a lifecycle table that clearly signals when to load each reference and concrete CLI entry points per phase. Its main gaps are the absence of validation checkpoints for the batch export/migrate workflow in the body and slight redundancy in the overview sections.
Suggestions
Add an explicit validation/verification checkpoint to the export/migrate row of the lifecycle table (e.g., after create-space, verify the imported Agent answers a sample question identically to the source before cutover), since batch/cross-workspace migration currently sequences get-space → remap → create-space with no verification step.
State the Genie-Agent-vs-Genie-One distinction once (either the Overview or the Related Skills section) instead of twice, and drop the opening line that duplicates the frontmatter description.
Include one copy-paste-ready CLI example for the most common action (e.g., a full `databricks genie create-space ...` invocation) in the body so the primary task is executable without first loading a reference file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean router with a dense, useful lifecycle table, but has minor trimmable redundancy: the Genie-Agent-vs-Genie-One distinction is stated in both the Overview and Related Skills sections, and the opening line repeats the frontmatter description. | 4 / 5 |
Actionability | The lifecycle table gives concrete CLI tokens ("discover-schema", "get-space --include-serialized-space", "genie-create-eval-run") and gating instructions ("Always load before creating or updating", "gather space ID + failing question + observed behavior first"), but full copy-paste-ready command syntax is deferred to the reference files. | 4 / 5 |
Workflow Clarity | The typical flow (create → query/validate → diagnose → optimize) and per-phase gating are clear, but the export/migrate phase — which explicitly covers batch and cross-workspace migration — sequences "get-space → remap → create-space" with no validation or verification step in the body, so the batch-operation cap applies. | 3 / 5 |
Progressive Disclosure | Exemplary structure: the lifecycle table maps each phase to a real reference file with explicit "Load when" conditions, all seven referenced paths exist on disk one level deep, and sibling cross-links between references serve only as signposting. | 5 / 5 |
Total | 16 / 20 Passed |