policy-comparator

# Policy comparator skill When asked to compare policy documents: 1. Use `read_file` to read the baseline policy (your organization's current policy or the standard you are comparing against) 2. Use `read_file` to read each comparison document 3. Identify the key policy dimensions by scanning all documents: - Coverage areas (e.g., eligibility, duration, benefits, exceptions) - Quantitative terms (e.g., days, percentages, dollar amounts) - Qualitative terms (e.g., approval process, escalation path) 4. For each dimension, classify each document's position relative to the baseline: - **Above**: more generous or comprehensive than baseline - **At**: equivalent to baseline - **Below**: less generous or comprehensive than baseline - **N/A**: dimension not addressed in this document 5. Produce the output as a comparison matrix: | Dimension | Baseline | Doc A | Doc B | Doc C | |-----------|----------|-------|-------|-------| | Eligibility | All FTE | Above | At | Below | 6. After the matrix, provide: - A summary paragraph highlighting where the baseline is strongest and weakest relative to the comparison set - Specific recommendations for policy updates, if requested 7. Use `write_file` to save the comparison if requested ## Important guidelines - Be precise about quantitative differences — "12 weeks vs 16 weeks" not just "below" - When a document is ambiguous about a dimension, quote the relevant text and mark it as "Unclear" rather than guessing - Do not make value judgments about which policy is "better" unless explicitly asked — present the facts and let the analyst decide - If documents are in different formats (PDF text, DOCX, markdown), normalize the comparison — format should not affect the analysis