metrics

Skill: Metrics

Purpose

Browse, search, and display metric definitions from the active dataset's metric dictionary. Provides quick access to how metrics are defined, computed, and validated.

When to Use

• User says /metrics or "show me the metrics" or "what metrics do we track?"
• During analysis, to confirm a metric's definition before computing it
• When writing a metric spec, to check for existing definitions

Invocation

/metrics — list all metrics for the active dataset /metrics {id} — show full spec for a specific metric /metrics category={cat} — filter by category (e.g., monetization) /metrics search={term} — search metric names and descriptions

Instructions

Step 1: Load Metric Dictionary

1. Read .knowledge/active.yaml to identify the active dataset.
2. Read .knowledge/datasets/{active}/metrics/index.yaml for the metric list.
3. If no metrics directory exists: "No metric dictionary for this dataset. Use the metric-spec skill to define metrics."

Step 2: Execute Command

List all (/metrics):

• Display as a table: id, name, category, direction, validation_status
• Group by category
• Show total count
• If dictionary is empty AND user mentions specific analysis context (e.g., "revenue analysis", "conversion analysis"):
  • Read .knowledge/datasets/{active}/schema.md to explore available tables/columns
  • Suggest 3-5 relevant metrics the user could define for their analysis context
  • Include suggested SQL formulas for each
  • Reference the metric-spec skill for formalization

Show specific (/metrics {id}):

• Read .knowledge/datasets/{active}/metrics/{id}.yaml
• Display: name, category, owner, full definition (formula, unit, direction, granularity), source tables, dimensions, guardrails, typical range, validation status
• If metric not found:
  • Suggest closest match from index (fuzzy string match on name)
  • Fallback search strategy: Search working/, outputs/, and .knowledge/analyses/ for recent usage of the metric name
  • If found in recent work, extract the formula/definition used and offer to formalize it
  • If not found anywhere, suggest defining it via metric-spec skill