Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a strong, well-structured advisory skill that provides clear workflows branching by context type, concrete MCP tool usage, and specific decision criteria for metric selection. Its main weakness is moderate verbosity — some explanatory prose and philosophical guidance could be trimmed without losing clarity. The important context section is excellent, surfacing non-obvious gotchas (CUPED/percentile incompatibility, context kind mismatches) that would otherwise cause real problems.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is well-structured but somewhat verbose in places — the explanatory prose around each context type (experiment, guarded rollout, release policy) could be tightened. Some guidance like 'Guarded rollouts are safety mechanisms, not experiments' and 'The discipline of choosing is the point' are stylistic rather than instructional. However, it largely avoids explaining things Claude already knows and most content earns its place. | 2 / 3 |
Actionability | The skill provides specific MCP tool calls (list-metrics, list-metric-events, list-release-policies), concrete output formats to display to users, explicit criteria for metric selection (e.g., 'Only recommend metrics with events actively flowing'), and typed recommendation categories with examples. The guidance is specific enough to execute without ambiguity. | 3 / 3 |
Workflow Clarity | The 5-step workflow is clearly sequenced with explicit branching by context type (experiment/guarded rollout/release policy). Step 2 has a clear skip condition for experiments. Step 3 includes a health-check validation gate before recommendations. The 'Important Context' section serves as a validation checklist covering edge cases like mid-experiment changes, CUPED incompatibility, and context kind mismatches. | 3 / 3 |
Progressive Disclosure | The skill is appropriately self-contained as an advisory document, with clear one-level-deep references to related skills (metric-create, metric-instrument) at the end. Content is well-organized with headers, tables, and collapsible context sections. The document doesn't try to inline creation or instrumentation workflows — it defers those cleanly. | 3 / 3 |
Total | 11 / 12 Passed |