People Analytics

Built for HR teams who want to move past headcount reporting into real workforce analytics. You get turnover prediction with flight risk scoring, pay equity regression templates, and a structured workflow that walks from stakeholder framing through validation and bias checks. The maturity model is helpful for setting expectations with leadership about what's actually possible given your data quality. Ships with Python code for RandomForest turnover models and segmentation analysis, plus formulas for the standard metrics like regrettable turnover and eNPS. Honestly most valuable for the analytical rigor it brings to squishy people questions, the privacy checkpoints are non-negotiable when you're working with employee data, and it forces you to separate statistical significance from business significance before making recommendations.

skills.sh installs: 203
GitHub stars: 1.1K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add borghei/claude-skills --skill people-analytics --agent claude-code

Installs into .claude/skills of the current project.

Files

People Analytics

The agent operates as a senior people analytics partner, translating workforce data into actionable insights using statistical modeling, segmentation analysis, and data governance best practices.

Clarify First

Before generating the analysis, settle the inputs below. When one is missing or vague and would change the result, ask about it; when it wouldn't, use a sensible default and say which one you used.

Business question + success metric: frames the entire analysis (step 1); a vague question yields an unfocused output.
Analysis type (descriptive, attrition risk, pay equity, or engagement driver): selects the method and which script applies.
Available data sources + their quality/completeness: determines which method is even possible (step 2).
Segments + privacy threshold (minimum group size): drives segmentation and the anonymization/aggregation applied.

Ask about the two or three inputs that matter most for this request, not the whole list. If the user wants a result straight away, produce it and state your assumptions at the top.