Content
38%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 has a reasonable skeleton — clear use/don't-use sections, a four-step instruction sequence, and safety guardrails — but delivers almost no actionable specifics, duplicates its own opening in a Purpose section, and points to 12 sub-skill files that are missing from the bundle entirely. It reads as a stub or index rather than a functioning skill.
Suggestions
Ship the ./sub-skills/*.md files referenced by the 12 knowledge-module links (or remove the links) — as delivered, every reference in the skill is broken.
Make the Instructions actionable: add concrete decision guidance or examples, e.g. naming when to pick Spark vs. dbt vs. Airflow, a sample dbt project layout, or an Airflow DAG skeleton for a batch pipeline.
Delete the redundant "## Purpose" section (it restates the opening line and frontmatter description) and the empty "## Capabilities" header; fold the production-write validation into the step sequence as an explicit checkpoint (validate → fix → re-validate before writing to sinks).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly tight bullets, but the "## Purpose" section ("Expert data engineer specializing in building robust, scalable data pipelines...") restates the opening line and the frontmatter description almost verbatim — a fully redundant padded section. Anchor 3 ("mostly efficient but some unnecessary explanation") rather than 2, since it is one padded section rather than several. | 3 / 5 |
Actionability | The Instructions are high-level hints — "Choose architecture, storage, and orchestration tools", "Monitor quality, costs, and operational reliability" — with no concrete commands, configuration examples, decision criteria, or code. Anchor 2 ("high-level hints but missing the specific steps to execute") rather than 1 because a real four-step approach is prescribed, and rather than 3 because no executable concrete guidance appears anywhere. | 2 / 5 |
Workflow Clarity | A clear four-step sequence exists ("Define sources, SLAs, and data contracts" → "Choose architecture..." → "Implement ingestion, transformation, and validation" → "Monitor quality, costs..."), and Safety notes "Validate data before writing to production sinks", but validation checkpoints are implicit rather than woven into the flow and there is no validate→fix→retry feedback loop for production-sink writes. Per the rubric's scoring note, missing feedback loops for batch/destructive operations caps this at 3. | 3 / 5 |
Progressive Disclosure | The 12 knowledge-module links ("### 1. [Modern Data Stack & Architecture](./sub-skills/modern-data-stack-architecture.md)" etc.) are clearly signaled and one level deep, but the ./sub-skills/ directory does not exist in the bundle — every reference is a dead end, so the disclosure structure does not actually function. There is also a stray empty "## Capabilities" header immediately before "## 🧠 Knowledge Modules". Scored against the actual bundle structure per the rubric guideline; anchor 2 rather than 3 because broken references leave the skill an empty shell, worse than merely under-signaled ones. | 2 / 5 |
Total | 10 / 20 Passed |