Content
92%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-engineered skill body: executable commands, explicit decision gates, validation feedback loops, and a clean one-level reference structure. The only weakness is mild redundancy across the routing/guidance sections that slightly inflates token cost.
Suggestions
Consolidate the Data Access Decision Gate, 'When to Use What', and State Storage Guidance sections to remove overlapping routing guidance and reduce token cost.
Tighten the long 'Is this even a Databricks App?' paragraph into a short bullet list of app-vs-AI/BI triggers.
Move the repeated 'derive --set keys from databricks apps manifest' instruction to a single canonical location and reference it.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly lean operational guidance (naming limits, anti-patterns, gotchas) with executable commands, but there is some repetition across the Decision Gate, 'When to Use What', and State Storage sections and a few long prose passages that could be trimmed, keeping it just below a 5. | 4 / 5 |
Actionability | Provides copy-paste-ready CLI commands (manifest, init with full --set examples, npx docs) plus explicit wrong/right anti-patterns, fully covering the common scaffolding and validation cases. | 5 / 5 |
Workflow Clarity | Sequenced phases ('FOLLOW THIS ORDER'), a required Data Access Decision Gate, a numbered Analytics workflow with a typegen-before-UI checkpoint, and explicit validate->fix->revalidate feedback loops for destructive/batch operations. | 5 / 5 |
Progressive Disclosure | Overview body routes to 15 real, one-level-deep reference files via a phase-organized table, with all referenced paths resolving to actual bundle files and clear navigation. | 5 / 5 |
Total | 19 / 20 Passed |