principle-synthesizer_skill

Overview

This skill synthesizes invariant principles from three or more sources to surface candidate Golden Masters. It identifies core ideas that survive across varied expressions while remaining transparent about methodology and limits. Use it to produce concise, evidence-backed principles for documentation or decision-making.

How this skill works

The skill gathers normalized extractions or raw texts, aligns candidate principles across inputs, and validates semantic consistency rather than simple keyword overlap. It classifies findings as invariant, domain-specific, or noise and outputs Golden Master candidates with supporting evidence and normalization metadata.

When to use it

• You have 3+ texts or extraction outputs and need the shared core principles
• You want a Golden Master candidate to serve as a canonical statement
• You need to filter noise and surface consistently expressed values or rules
• You want evidence-backed principles for documentation or training
• You must compare 3–7 sources to validate recurring concepts

Best practices

• Provide at least 3 sources; 3–7 is recommended for clear results
• Supply normalized extractions when available to reduce re-normalization
• Keep sources within a single domain to avoid incompatible comparisons
• Review Golden Master candidates manually before adopting as canonical
• Track original variant phrasing to preserve voice and provenance

Example use cases

• Distill company values from multiple internal documents into canonical statements
• Create canonical design principles from several product guidelines
• Synthesize recurring ethics statements found across policy drafts
• Compress repeated recommendations in technical specs into a small set of invariants
• Identify persistent user needs across multiple research transcripts