Content
72%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a concise, well-organized reference of SQL best practices that respects Claude's existing knowledge and stays token-efficient. Its main gaps are the absence of any complete runnable example query and the lack of an explicit validation/feedback loop for destructive or batch database operations.
Suggestions
Add one or two complete, executable example queries (e.g., a keyset-pagination query and an EXPLAIN-verified optimization before/after) to lift actionability to fully copy-paste ready.
For risky operations (schema migrations, bulk writes, adding indexes), include an explicit validate→fix→retry loop, e.g., run EXPLAIN ANALYZE before/after, test on a copy, and verify row counts before committing.
Consider splitting the optimization and schema sections into referenced detail files only if the content grows; at its current size the inline organization is already appropriate.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean bullet guidance that assumes Claude already knows SQL, with parentheticals only for genuinely non-obvious points (CTE materialization, function-on-index blocking index use). Not 2 because there is no padding or unnecessary concept explanation; every bullet earns its place. | 3 / 3 |
Actionability | Directives are concrete and specific ("Use EXISTS instead of IN," "WHERE id > last_seen_id," "TIMESTAMP WITH TIME ZONE," anti-pattern "WHERE YEAR(created_at) = 2025"), but there are no complete executable example queries or copy-paste snippets. Not 3 because no runnable code block is provided; not 1 because the guidance is specific and actionable rather than abstract. | 2 / 3 |
Workflow Clarity | Content is well organized by category and the "Pitfalls to Avoid" section plus "use EXPLAIN or EXPLAIN ANALYZE" and "Do not add indexes without measuring" act as implicit checkpoints, but there is no explicit validate→fix→retry feedback loop for risky write/batch operations (migrations, indexing, bulk updates), which the notes say should cap this at 2. Not 1 because sequencing/checkpoints are partly present; not 3 because no explicit feedback loop is defined. | 2 / 3 |
Progressive Disclosure | The skill is under 50 lines, single-purpose with no bundle files, and is organized into clear labeled sections (Key Principles, Query Optimization, Schema Design, Analysis Patterns, Pitfalls). Per the guideline, this can score 3 on well-organized sections alone; not 2 because the structure is clean and there are no nested references. | 3 / 3 |
Total | 10 / 12 Passed |