Content
40%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 SKILL.md body is concise and well-organized as an overview, but it is fundamentally a router whose destination files are missing: the content map points to six reference .md files that do not exist in the bundle, leaving the actual actionable and workflow detail unavailable. What remains is high-level principle and anti-pattern guidance with no executable steps and no validation checkpoints.
Suggestions
Add the missing reference files (database-selection.md, orm-selection.md, schema-design.md, indexing.md, optimization.md, migrations.md) to references/ so the content map actually resolves, or inline the essential decision guidance directly into SKILL.md.
Provide concrete, executable decision criteria or code instead of high-level directives — e.g. specific "if serverless / if simple app → choose X" rules or example migration commands.
Add a validation/feedback checkpoint to the workflow for destructive operations, e.g. "After drafting a migration, verify it is reversible and run it against a copy of the schema before applying."
Either reference scripts/schema_validator.py where relevant or remove it, since it is currently an orphaned bundle file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean — bullet points, a content-map table, a checklist, and an anti-patterns list — and assumes Claude already knows database concepts, with only minor trimmable noise such as the "Learn to THINK" quote and the generic "When to Use" boilerplate. | 4 / 5 |
Actionability | Guidance is high-level ("Choose database/ORM based on CONTEXT", "ASK user for database preferences") with a few heuristics ("SQLite may suffice"), but it delegates the actual executable detail to the reference .md files in the content map — and those files do not exist in the bundle. | 2 / 5 |
Workflow Clarity | A pre-design "Decision Checklist" and the selective-reading rule give a rough sequence, but there is no sequenced workflow for schema design or migrations and no validation/feedback checkpoints, which the rubric requires for destructive database operations. | 2 / 5 |
Progressive Disclosure | The SKILL.md presents a clean, well-signaled one-level-deep content map, but every referenced file (database-selection.md, orm-selection.md, schema-design.md, indexing.md, optimization.md, migrations.md) is absent — the references/ directory does not exist — so the disclosure mechanism is broken and Claude cannot retrieve the promised detail content; the lone bundle file, scripts/schema_validator.py, is never referenced. | 2 / 5 |
Total | 10 / 20 Passed |