Content
92%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.
An exceptionally lean and expert-level review-lens skill that assumes Claude's competence throughout. The ordered method, explicit thresholds, and copy-paste fix commands make it immediately actionable; the only minor gap is telling the reviewer to supply sample verification SQL without offering one itself.
Suggestions
Include one short sample read-only verification query in the Reporting section (e.g., a grouped mapping-count SELECT) so the 'give sample read-only SQL' instruction has a concrete model to follow.
The Scope bullet 'Transactions and indexes' bundles two distinct checks; splitting it into transaction scope and online index creation would make the checklist marginally easier to scan.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Phrases like 'The dump's version stamp should match the change's newest migration' and 'The fix is expand and contract: add, migrate readers and writers, remove only once nothing reads the old shape' deliver dense, non-redundant expertise with zero padding or explanation of concepts Claude already knows. Every line is either a check, a method step, a threshold rule, or a reporting instruction. | 5 / 5 |
Actionability | The body gives concrete, executable commands ('git checkout <base> -- db/schema.rb', 'rerun bin/rails db:migrate', 'Diff the dump and db/migrate/ against the review's base ref, never an assumed main') plus specific inspection techniques ('Check each CASE or IF branch and constant hash entry individually'). It falls short of 5 because the reporting section says 'give sample read-only SQL that would prove it' without providing a sample, and not 3 because the guidance is directly executable as written, not pseudocode. | 4 / 5 |
Workflow Clarity | The Method section gives an explicit ordered sequence ('check drift first... Then check migration correctness, then verification and rollback') with a guardrail ('Do not trust fixtures; production data shapes differ'), and the Threshold section acts as a report/don't-report checklist. This is a single-purpose, read-only review skill under 50 lines, so the simple-skill exception applies and no destructive-operation cap is triggered since the skill only reviews, never performs, destructive changes. | 5 / 5 |
Progressive Disclosure | The skill is under 50 lines with no need for external references (no references/, scripts/, or assets/ bundle exists), and its content is organized into clear, well-labeled sections (Scope, Method, Threshold, Reporting). Per the rubric guideline, this earns a 5 with just well-organized sections. | 5 / 5 |
Total | 19 / 20 Passed |