Content
75%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.
A highly actionable, well-structured skill: every workflow step is executable with concrete CLI commands, parameters, and naming rules verified against the bundled script. Its main weakness is token efficiency — a substantial section re-teaches standard SQL that Claude already knows, and examples repeat command syntax redundantly.
Suggestions
Cut the 'Analysis Patterns' section down to only DuckDB-specific or non-obvious patterns (e.g., cross-file joins, name sanitization/collisions), removing basic SQL (COUNT, DISTINCT, GROUP BY, window function syntax) that Claude already knows.
Condense the 'Complete Example' and 'Multi-file Example' to just the SQL and novel flags, since the analyze.py invocation syntax is already fully specified in the Workflow and Parameters sections.
Add a brief error-recovery note (e.g., 'if a query fails on an unknown table, re-run --action inspect and use the printed SQL names') to close the validation-gap in workflow clarity.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 'Analysis Patterns' section (~55 lines) teaches SQL Claude already knows (COUNT/DISTINCT, GROUP BY, window functions, CASE pivots), and the 'Complete Example' repeats identical CLI invocations already shown. However, there is no padded concept explanation and the bulk is executable commands, so it fits 'mostly efficient but includes some unnecessary explanation'. | 3 / 5 |
Actionability | Every command is copy-paste ready with real flags (--files, --action, --sql, --table, --output-file), backed by a parameter table, output-format list, table-naming rules, and end-to-end complete examples covering the common cases. | 5 / 5 |
Workflow Clarity | Clear sequence (understand requirements → inspect → query/summary/export) with inspect serving as an implicit checkpoint ('Use the SQL names shown by inspect'). Operations are read-only so the destructive-cap does not apply, but there is no explicit error-recovery/feedback loop (e.g., what to do when a table name or sheet is not found), keeping it below 5. | 4 / 5 |
Progressive Disclosure | Well-sectioned single-script skill with the bundle script correctly referenced and deliberately kept opaque ('Do NOT read the Python file, just call it with the parameters'). At ~255 lines all inline, the generic SQL patterns and extended examples could be split into a one-level-deep reference file, so it is good-but-not-ideal structure. | 4 / 5 |
Total | 16 / 20 Passed |