Content
62%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured process skill with strong workflow clarity and a thoughtful critique framework, but it falls short on actionability by not including any concrete SQL examples despite being a query-writing skill. The content is moderately concise — it communicates the right ideas but uses more tokens than necessary, particularly in explanatory sections. The skill would benefit significantly from at least 2-3 complete, executable BQ query templates and from splitting reference material (table schemas, metric definitions) into separate files.
Suggestions
Add 2-3 complete, executable BigQuery SQL query examples (e.g., a funnel conversion query segmented by platform, a time-trend query) rather than just describing query approaches in prose.
Extract Headout-specific reference material (BQ table names/schemas, metric definitions like S2O/C2O/GBV, standard segmentation dimensions) into a separate REFERENCE.md file to reduce SKILL.md length and improve progressive disclosure.
Tighten the structured critique section — convert the paragraph-length explanations of each bias into 1-2 line checklist items with a brief 'check: [specific thing to verify]' format.
Remove or compress the introductory framing paragraph ('You are the Data Analyst specialist...') — Claude doesn't need role-setting prose when the skill structure already makes the role clear.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is reasonably well-written but includes some unnecessary framing ('You are the Data Analyst specialist...', 'Think like an analyst who has seen a thousand funnels') and explanatory padding that Claude doesn't need. The structured critique section, while valuable, is verbose — each bias check could be tightened. The standards section restates what's already implied. Overall functional but could lose ~30% of tokens without losing information. | 2 / 3 |
Actionability | The skill provides good structural guidance (question dimensions, output templates, critique checklist) but lacks concrete executable code. No actual SQL queries are provided despite being a query-focused skill — only vague descriptions like 'Segment S2O by variant_count bucket.' The example section shows question bank output but not a real BQ query. For a skill centered on writing and running BigQuery queries, the absence of even one complete SQL example is a significant gap. | 2 / 3 |
Workflow Clarity | The workflow is clearly sequenced: Step 1 (load context) → Step 1.5 (surface blind spots with explicit completion criteria) → Mode A (question bank) → Mode B (query runner) → Structured critique → Output. Each mode has clear entry conditions ('when to use'), the blind-spot step has an explicit 'complete when' gate, and the structured critique serves as a validation checkpoint before finalizing output. The feedback loop of 'present findings to PM before finalising' is explicit. | 3 / 3 |
Progressive Disclosure | The skill references external files (CLAUDE.md, Problem Frame docs) appropriately, but the skill itself is a monolithic ~300-line document with no references to supporting files for detailed content like BQ table schemas, metric definitions, or example queries. The common tables section and metric definitions could be split into reference files. With no bundle files provided, the skill tries to be self-contained but becomes long as a result. | 2 / 3 |
Total | 9 / 12 Passed |