Content
75%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.
The body is highly actionable: complete executable YAML for both dbt variants, a concrete before/after example, and precise Orchestra-specific gotchas with exact syntax for semicolon joining, env-var naming, and MANUAL flags. Its main weaknesses are redundancy — the project_dir and package_manager rules are each restated three or four times across the table, steps, and gotchas — and the absence of any bundle references despite a body over 150 lines, leaving the linked-connections detail inline.
Suggestions
State each rule once: keep the Parameter Mapping table as the single source for project_dir and package_manager guidance, and trim the verbatim repeats in Conversion Steps and Gotchas to short cross-references.
Move the 'Warehouse Credentials: Link, Don't Copy' section (env-var naming scheme, profiles.yml example, failure modes) into a references/ file (e.g. references/linked-connections.md) linked from a brief summary, reducing SKILL.md toward an overview.
Add an explicit final validation step to Conversion Steps, e.g. 'Render the YAML and confirm the trailing semicolon, connection name, and any # MANUAL: flags with the user before considering the migration done.'
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with Orchestra-specific facts Claude would not know (integration names, env-var naming scheme, linked-connection behavior) and does not explain dbt/Prefect basics. However, the same rules are repeated across sections — project_dir guidance appears in the mapping table, Conversion Steps, Gotchas, and comments in both YAML examples; package_manager inference appears three times — which is more than minor over-explanation, fitting anchor 3 rather than 4. | 3 / 5 |
Actionability | Provides two complete copy-paste YAML examples (dbt Core and dbt Cloud), a full before/after Python-to-YAML pair, an exact '# MANUAL:' comment pattern for the ambiguous package_manager case, and an executable env_var() profiles.yml snippet. Fully executable and covering the common cases, matching anchor 5. | 5 / 5 |
Workflow Clarity | Conversion Steps is a clearly sequenced checklist with variant branching (dbt Core vs dbt Cloud) and flag-don't-guess checkpoints for package_manager and unsupported Python versions. There is no explicit post-conversion validation step (e.g. confirming the YAML or reviewing output with the user), so it fits anchor 4 rather than 5; the task is code generation, not destructive/batch, so the workflow-clarity cap at 3 does not apply. | 4 / 5 |
Progressive Disclosure | No bundle files exist, so everything is inline in a ~170-line body. Section headers are clear, external doc links are grouped under References, and alert handling is appropriately offloaded to the prefect-alerts-to-orchestra skill. Because the body exceeds 50 lines and all content (including the detailed linked-connections section) is inline with no well-signaled reference files, it fits anchor 4 rather than 5. | 4 / 5 |
Total | 16 / 20 Passed |