Content
100%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a lean dispatch layer with a clear, validated workflow and one-level-deep references that all resolve to real bundle files. It exemplifies progressive disclosure and assumes Claude's intelligence throughout.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body assumes Claude's competence and avoids restating basic concepts (what an experiment is, what SRM means in the abstract); every line is dispatch or state-calibration guidance that earns its place, with detail deferred to reference files. | 3 / 3 |
Actionability | Names concrete executable inputs — the exact `experiment-get` fields to pull, the `multiple_variant_handling` default, the diagnostic-snapshot MCP tools to run — and points to specific reference files with verified paths, giving copy-paste-ready direction. | 3 / 3 |
Workflow Clarity | A clear Step 1 → 1.5 → 2 → 3 → 4 sequence with an explicit verification checkpoint ('Pull a diagnostic snapshot — verify before asking') and a fallback loop ('If the symptom is unclear, ask one clarifying question'); diagnostics carry HIGH/MEDIUM/LOW verification tags. | 3 / 3 |
Progressive Disclosure | SKILL.md is an overview/dispatch table pointing one level deep to five real reference files (bias-and-skew.md, empty-experiment.md, interpretation.md, numbers-vs-sql.md, mid-run-changes.md) and diagnostic-snapshot.md, each confirmed present; navigation is clearly signaled and content appropriately split. | 3 / 3 |
Total | 12 / 12 Passed |