Content
10%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 a persona-style capability catalogue rather than actionable skill guidance: it pads the context window with tool lists Claude already knows while providing no executable steps, no validation checkpoints, and no reference structure. It needs to be replaced with concise, task-specific instructions (with code/commands where relevant) and offload capability catalogues to reference files.
Suggestions
Cut the 'Capabilities', 'Knowledge Base', and 'Behavioral Traits' catalogues — Claude already knows what Spark, Airflow, dbt, and Kafka are; keep only non-obvious guidance (e.g. project-specific conventions, chosen stack, naming standards).
Replace the abstract 'Instructions'/'Response Approach' lists with concrete, executable guidance: example pipeline code, dbt model skeletons, or specific design checklists for the common request types listed in 'Example Interactions'.
If detailed capability/per-platform detail must be kept, move it into a references/ file (one level deep, clearly linked from SKILL.md) instead of inlining ~180 lines in the main file, and add explicit validation checkpoints (e.g. 'run data quality checks before loading to production') to the workflow steps.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly 180 of ~215 lines are tool catalogues ('Relational databases: PostgreSQL, MySQL, SQL Server integration', 'Data lakes: AWS S3, Azure Data Lake, Google Cloud Storage...') enumerating technologies and capabilities Claude already knows. This is severely padded content that adds almost nothing Claude does not already know, matching the lowest anchor. | 1 / 5 |
Actionability | The instructions ('Define sources, SLAs, and data contracts', 'Choose architecture, storage, and orchestration tools') are entirely abstract with no code, commands, or executable steps — the body only describes rather than instructs, matching the 'entirely vague or abstract' anchor. | 1 / 5 |
Workflow Clarity | Two step lists exist ('Instructions' 1–4, 'Response Approach' 1–8) but the steps are loosely defined and validation appears only as a Safety principle ('Validate data before writing to production sinks'), never as a checkpoint in the actual sequence. Pipeline builds writing to production sinks are batch operations, so missing feedback loops cap this at 3; the rough, gap-filled sequence lands it at 2. | 2 / 5 |
Progressive Disclosure | There are no bundle files at all, and ~180 lines of capability/reference-style catalogue content that clearly belongs in separate files are inlined in SKILL.md. Section headers exist (so not a 1), but the structure is minimal and nothing is split out, matching the 'content that clearly belongs in separate files is inlined' anchor. | 2 / 5 |
Total | 6 / 20 Passed |