Content
68%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a highly actionable and reasonably concise Snowflake reference that excels at providing executable code examples and highlighting common pitfalls (colon prefix, TO_FILE args, deprecated functions). Its main weaknesses are the monolithic structure — covering 8+ major topics in a single file without any external references — and the lack of explicit validation/feedback loops for pipeline and database operations.
Suggestions
Split into separate referenced files (e.g., CORTEX_AI.md, PIPELINES.md, PERFORMANCE.md) and keep SKILL.md as a concise overview with signposted links to each topic.
Add explicit validation steps for pipeline workflows, e.g., 'After creating a Dynamic Table, verify refresh status: SELECT * FROM TABLE(INFORMATION_SCHEMA.DYNAMIC_TABLE_REFRESH_HISTORY(...))' and 'After resuming a task, confirm execution: SELECT * FROM TABLE(INFORMATION_SCHEMA.TASK_HISTORY(...))'.
Remove the 'When to Use' section — it restates the skill description and Claude can infer applicability from the content itself.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The content is largely efficient, providing Snowflake-specific knowledge Claude wouldn't inherently have (colon prefix rule, TO_FILE argument separation, deprecated functions list). Minor verbosity in the 'When to Use' section and a few explanatory asides (e.g., 'Snowflake's columnar storage scans only referenced columns') that Claude likely knows, but overall well-trimmed for the breadth covered. | 4 / 5 |
Actionability | Nearly every section includes executable, copy-paste-ready SQL or Python code covering common cases. The BAD/GOOD pattern comparisons for stored procedures and TO_FILE are particularly actionable, and the error patterns table maps directly from symptom to fix. | 5 / 5 |
Workflow Clarity | The Streams+Tasks section shows a clear sequence including the critical 'Tasks start SUSPENDED — you MUST resume them' checkpoint. However, for a skill covering data pipelines and database operations (destructive/batch contexts), there are no explicit validation/feedback loops — no 'verify the Dynamic Table is refreshing correctly' or 'check stream offset after task execution' steps. Per rubric rules, missing feedback loops for database operations caps this at 3. | 3 / 5 |
Progressive Disclosure | This is a very large monolithic file (~250+ lines) covering SQL, pipelines, Cortex AI, Cortex Agents, Snowpark, dbt, performance, and security — all inlined with no references to separate files. Given the breadth, sections like Cortex AI function reference, Cortex Agents spec format, and the error patterns table would benefit from being split into separate referenced files. No bundle files exist to support this. | 2 / 5 |
Total | 14 / 20 Passed |