European-Parliament-MCP-Server: Skill for Claude Code

Data Science for Parliamentary Intelligence

Context

This skill applies when:

• Performing statistical analysis of MEP voting patterns using EP MCP Server data
• Applying NLP techniques to EU legislative texts, parliamentary questions, and committee reports
• Building network models of political group alliances and MEP collaboration patterns
• Conducting time series analysis of voting cohesion, attendance, and legislative productivity
• Using dimensionality reduction to map MEP ideological positions from roll-call votes
• Clustering MEPs by policy area specialization and voting behavior
• Forecasting legislative outcomes based on committee vote patterns and amendment analysis
• Detecting anomalies in MEP activity patterns (unusual voting, attendance drops)

All data science activities must comply with Hack23 ISMS policies for data integrity, GDPR-compliant processing, and reproducible analysis.

Rules

1. Data Quality First: Validate all EP MCP Server data before analysis — check for missing values, encoding issues in multilingual text, and temporal gaps in voting records
2. Appropriate Methods: Match analytical technique to data type — nominal (political group), ordinal (agreement scale), interval (vote margins), network (co-sponsorship)
3. Reproducibility: Document all data transformations, model parameters, and analytical decisions — analysis must be reproducible from MCP Server queries
4. Statistical Rigor: Report effect sizes, confidence intervals, and p-values — avoid p-hacking by pre-registering hypotheses before querying EP data
5. GDPR-Aware Processing: When analyzing MEP personal data, apply data minimization — aggregate where possible, pseudonymize when individual tracking is not analytically necessary
6. Bias Awareness: Account for selection bias in roll-call votes (only ~30% of EP votes are roll-call), survivorship bias (only re-elected MEPs span terms), and measurement bias (attendance ≠ engagement)
7. Domain Validation: Validate statistical findings against political science domain knowledge — a statistically significant pattern must also be institutionally plausible
8. Scalable Pipelines: Design analysis pipelines that can process data across legislative terms (EP5–EP10) and scale to 720+ MEPs per term
9. Visualization Standards: Use clear, accessible visualizations — label axes, include legends, use colorblind-safe palettes, and annotate key political events on time series
10. Ethical AI: If applying ML models to MEP behavior prediction, disclose model limitations, avoid deterministic predictions about individual MEPs, and never use models for discriminatory profiling