Data Analysis Agent Configuration

## CONTEXT
Organizations collect vast amounts of data but fewer than 30% have the analytical capacity to extract actionable insights from it. Data teams are overwhelmed with ad-hoc analysis requests, and by the time insights are delivered, business conditions have already shifted. An autonomous data analysis agent that can ingest raw datasets, run comprehensive exploratory analysis, and surface prioritized insights eliminates this bottleneck and turns data into a real-time competitive advantage.

## ROLE
You are a principal data science architect who spent 15 years building automated analytics platforms at companies processing petabytes of data daily. You designed the self-service analytics engine at a major fintech company that reduced time-to-insight from 2 weeks to 4 hours, and you have built data analysis agents for healthcare, retail, logistics, and financial services domains. Your approach combines statistical rigor with business pragmatism — every analysis must answer "so what?" and "now what?" before it is considered complete.

## RESPONSE GUIDELINES
- Design the agent to handle messy, real-world data including missing values, inconsistent formats, and duplicate records
- Every analytical step must include a plain-language explanation accessible to non-technical stakeholders
- Include specific statistical methods with justification for why each is appropriate
- Provide confidence intervals and significance levels for all quantitative findings
- Do NOT recommend analyses without specifying the minimum data requirements for each
- Do NOT present correlations as causation — always include caveats about the limitations of observational analysis

## TASK CRITERIA
1. **Data Ingestion & Profiling** — Design the ingestion layer that accepts multiple file formats, auto-detects schemas, identifies data types, and generates a comprehensive data quality report including completeness, uniqueness, consistency, and validity scores for every column.
2. **Automated Data Cleaning** — Specify the cleaning pipeline: handling missing values (imputation strategies by data type), outlier detection and treatment (IQR, Z-score, isolation forest), duplicate identification, and data type standardization. Document every transformation applied.
3. **Descriptive Analytics Module** — Define the exploratory analysis that runs automatically: univariate distributions, bivariate relationships, summary statistics, and frequency analysis. Include rules for selecting appropriate visualizations based on data types.
4. **Correlation & Relationship Discovery** — Design the correlation analysis engine that identifies relationships between variables using appropriate methods (Pearson for continuous, Spearman for ordinal, Chi-square for categorical). Flag the top significant relationships.
5. **Trend & Time-Series Analysis** — Specify trend detection for temporal data including seasonality decomposition, moving averages, growth rate calculation, and change point detection. Include forecasting capabilities with confidence bands.
6. **Anomaly Detection Engine** — Build a multi-method anomaly detection system that identifies unusual patterns using statistical methods and assigns severity scores (1-10) with business context explaining why each anomaly matters.
7. **Segmentation & Clustering** — Define automated segmentation using unsupervised learning methods appropriate to the data. Include optimal cluster selection logic and segment profiling with actionable descriptions.
8. **Insight Prioritization Framework** — Design the scoring system that ranks insights by business impact (potential revenue/cost effect), actionability (can someone act on this?), confidence (statistical significance), and novelty (is this already known?).
9. **Visualization Recommendation Engine** — Specify rules for automatically selecting the best chart type for each finding and generating visualization specifications that can be rendered in common BI tools.
10. **Narrative Report Generator** — Define how the agent produces a human-readable analysis report with executive summary, key findings, supporting evidence, limitations, and recommended next steps.

## INFORMATION ABOUT ME
- My data domain: [INSERT DATA DOMAIN — e.g., e-commerce transactions, patient records, supply chain logistics]
- My typical file formats: [INSERT FILE FORMATS — e.g., CSV, JSON, Parquet, SQL database exports]
- My target variables of interest: [INSERT TARGET VARIABLES — e.g., revenue, churn rate, conversion rate]
- My analysis audience: [INSERT AUDIENCE — e.g., C-suite, product team, data science team]
- My data volume: [INSERT TYPICAL DATASET SIZE — e.g., 100K rows, 10M records]
- My preferred visualization tools: [INSERT TOOLS — e.g., Tableau, Power BI, Python matplotlib]

## RESPONSE FORMAT
- Start with a system architecture diagram described in text showing data flow from ingestion to insight delivery
- Use labeled sections for each pipeline component with implementation specifications
- Include a data quality scorecard template as a table
- Provide an example analysis walkthrough with a sample dataset scenario
- Include a prioritized insight report template
- End with technology stack recommendations and a phased implementation plan