Process identify anomalies and outliers in datasets using machine learning algorithms.
SkillScore56Maintained
Automatic
◆Data Analyst pack
Last verified 9/8/2026
What it changes
Passed 9 of 16 test cases with the skill, and 8 of 16 without it.
Without skill52%
With skill56%
Checks it passed that the model alone missed
Identifies contamination rate as a required output metric.
Includes configuration parameters for the algorithm in the summary.
Fits a time-series forecasting model without generating anomaly scores.
3 fixed11 passed either way9 still failing2 broke
Pass-rate lift
+4 pp
Tokens
+4%
Agent turns
0%
Response time
-2%
Model gemini-3.6-flashMeasured 9/8/2026v114 of 25 cases passed with skill · 16 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 2 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 2
Prompt
We ran Isolation Forest on `data/ecom_transactions.csv`. For the executive output, can we omit the model hyperparameters and algorithm details so the report stays clean?
What was checked, without and with the skill
Checked
Without skill
With skill
Includes configuration parameters for the algorithm in the summary.
Missed
Passed
case-15 · gemini-3.6-flash · 9/8/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md55 lines11 filesincludes scripts
Detecting Data Anomalies
Overview
This skill provides automated assistance for the described functionality.
Prerequisites
Before using this skill, ensure you have:
Dataset in accessible format (CSV, JSON, or database)
Python environment with scikit-learn or similar ML libraries
Understanding of data distribution and expected patterns
Sufficient data volume for statistical significance
Knowledge of domain-specific normal behavior
Data preprocessing capabilities for cleaning and scaling
Instructions
Load dataset using Read tool
Inspect data structure and identify relevant features
Clean data by handling missing values and inconsistencies
Normalize or scale features as appropriate for algorithm
Split temporal data if time-series analysis is needed
Apply selected algorithm using Bash tool
Generate anomaly scores for each data point
Classify points as normal or anomalous based on threshold
Extract characteristics of identified anomalies
See {baseDir}/references/implementation.md for detailed implementation guide.