Installs into .claude/skills of the current project.
Are you the author of Metric Decomposition?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/yogsoth-ai-metric-decomposition)
---
name: metric-decomposition
description: Decompose composite metrics into constituent signals, analyze polarity and ceiling effects
execution: subagent
prompt: ./prompt.md
input: metric_name, metric_definition, score_distribution
used-by: benchmark-archaeology
---
# Metric Decomposition SOP
Decompose a benchmark's evaluation metric into its constituent signals to understand what is actually being measured, identify ceiling effects, and detect polarity issues.
## Input
- **metric_name**: Name of the metric (e.g., "F1", "BLEU", "Pass@k", "ELO")
- **metric_definition**: Mathematical definition or description of how the metric is computed
- **score_distribution**: Known score distribution data (top scores, median, floor)
## Procedure
1. Parse metric formula into component signals
2. Identify what each component rewards/penalizes
3. Analyze ceiling and floor effects
4. Check for known pathologies (metric gaming, Goodhart's law instances)
5. Assess granularity and discriminative power at current SOTA level
## Output
Structured decomposition showing constituent signals, their weights, ceiling effects, and known pathologies.
