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 with exact syntax, complete YAML, and a concrete before/after example, and its conversion checklist is clearly sequenced. Its main weaknesses are token duplication — the same ~40-line YAML pipeline appears twice nearly verbatim — and the lack of a final validation checkpoint. Cross-skill navigation to sibling skills and external docs is clearly signaled.
Suggestions
Delete one of the two near-identical YAML pipeline blocks — either keep the Before/After example (which shows both sides) and drop the 'Orchestra YAML Structure' section, or keep the structure section and trim the After example to only the changed fields.
Mention allow_failure=True → treat_failure_as_warning once (in the mapping table) and remove its repetition in both the Conversion Steps and the Gotchas list.
Add a final checklist step to verify generated conditions reference real task IDs and output keys, e.g., 'Confirm every condition's PIPELINE_RUN_TASKS task-id exists and its OUTPUTS key is set by an upstream set_outputs: true task'.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body avoids explaining concepts Claude already knows and uses tight table/gotcha formats, but the 'Orchestra YAML Structure' example and the 'After' half of the Before/After example are near-identical ~40-line YAML blocks (only the pipeline name differs), and allow_failure guidance appears in the mapping table, the conversion steps, and the gotchas. The duplication is more than minor trimming, keeping it below anchor 4 but well above anchor 2's padded-prose verbosity. | 3 / 5 |
Actionability | Fully executable guidance throughout: exact Orchestra field names, complete copy-paste-ready YAML pipelines, exact condition syntax ("${{ ORCHESTRA.PIPELINE_RUN_TASKS['task-check-new-data'].OUTPUTS['row_count'] > 0 }}"), and a concrete before/after pair covering the most common case. Deferrals (set_output client setup, matrix) are explicitly routed to a named sibling skill. | 5 / 5 |
Workflow Clarity | The Conversion Steps checklist is a clear, well-ordered sequence (identify branches, classify input-driven vs output-driven, write conditions, replace allow_failure/wait_for, flag .map()), and the classification step acts as an implicit checkpoint. The conversion is neither destructive nor batch, so no cap applies; the gap is the absence of any final validation step (e.g., verify conditions reference real task outputs), which is a minor rather than structural gap. | 4 / 5 |
Progressive Disclosure | Well-organized single-file skill with clear sections (mapping table, structure, steps, example, gotchas) and well-signaled one-level references: named sibling skills ('See prefect-data-passing-to-orchestra', 'See prefect-alerts-to-orchestra') and external doc links. No bundle files exist to split content into, but at ~165 lines the duplicated full YAML examples are inline content that better organization would consolidate, so it does not reach anchor 5. | 4 / 5 |
Total | 16 / 20 Passed |