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.
An exemplary skill body: fully executable commands and queries, a clearly sequenced route-selection workflow with explicit error-recovery loops, and a clean split between inline operational guidance and the one offloaded column reference. The only flaw is minor redundancy between the route comparison table and the Route B difference bullets.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with non-obvious, dbt-specific facts (misrouting symptoms, double-encoded config, 'null' string quirk, stale target/ dirs) and teaches nothing Claude already knows, so it mostly assumes competence. Minor trimming is possible — the Route A/B comparison table and the Route B 'How this differs' bullets repeat each other, and the Contents table re-lists the section headers — which keeps it below the every-token-earns-its-place anchor. | 4 / 5 |
Actionability | Everything is copy-paste executable: complete 'dbt show' commands with the right flags, six ready-to-run SQL queries (untested models, missing descriptions, recursive lineage, materializations, incremental strategies, slowest models), a check file plus dbt_project.yml and checks YAML, and even the describe query to discover columns. Specific error codes ('InfoSchemaUnavailable (dbt1656)', 'unknown function: Jinja macro or function ref is unknown') make failure diagnosis concrete. | 5 / 5 |
Workflow Clarity | The routing workflow is clearly sequenced (verify dbt v2 → confirm metadata exists via the invocations query → pick Route A/B via the trade-off table) with explicit validation and feedback loops: 'If you get 0 rows from a view that should have data, refresh the metadata with dbt parse --generate-info-schema and retry'; misrouted queries are diagnosed by symptom ('A slow query or a warehouse error means the query went to the wrong engine'); Route B staleness is checked by comparing file mtime against the invocations row; and the check-conversion workflow includes testing with 'dbt check <name> --select <resources>' and cautionary gating before creating files. | 5 / 5 |
Progressive Disclosure | The bulky column reference is correctly split into references/view-columns.md (a real file, one level deep, clearly signaled in the Reference section with a descriptive label), while the operational guidance (gotchas, common queries, checks) stays inline where it is needed at query time. A Contents routing table at the top plus a route-decision table makes navigation easy, and links are one level deep with no nesting. | 5 / 5 |
Total | 19 / 20 Passed |