Skip to content

Latest commit

 

History

History
 
 

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

README.md

GPS Research Methodology

Prompts for research workflows designed around the Genealogical Proof Standard (GPS). These are structured to follow professional methodology developed by the Board for Certification of Genealogists and documented in Mills' Evidence Explained.

Core Concepts

GPS ensures genealogical conclusions are well-reasoned and evidence-based through five interdependent elements:

  1. Reasonably exhaustive research
  2. Complete, accurate citations
  3. Thorough analysis and correlation
  4. Resolution of conflicting evidence
  5. Coherent written conclusion

Recommended Prompts

Prompt Purpose Use When
research-assistant-v8.5-compact.md GPS-aligned compact research assistant Recommended compact prompt
research-assistant-v8.5-compact-unwrapped.md Same v8.5.2c compact prompt without hard wrapping Copy/paste into tools that preserve line breaks poorly
research-assistant-v8.md Full GPS-based research assistant (669 lines) System prompt for ongoing research assistance
research-assistant-v8-compact.md Token-efficient v8 Context-limited models
research-with-citations-v7.md Web research with GPS methodology Single research queries with citation requirements
web-research-v7.md Compact web search prompt Quick web research tasks
contract-first-genealogy-v3.1.md Contract-lock workflow Structured research projects with deliverables
research-agent-assignment-v2.1.md Research agent specification Autonomous research tasks

Full vs. Compact Variants

The compact/compressed versions are token-efficient alternatives for context-limited models. Here's what they trade away:

Feature Full v8 (704 lines) Compact v8 (147 lines)
Three-Layer Model Detailed tables with examples Condensed paragraph summaries
Worked examples Death certificate walkthrough Omitted
Adaptive scaffolding Full detection protocol + calibration table Simplified behavioral cues
Decision trees Appendix B flowcharts Omitted
Core templates Evidence table, research plan, conflict matrix, proof summary Omitted
Terminology reference Appendix A quick-reference table Omitted
Provenance chain Detailed with error propagation Brief mention
Cluster research Full FAN protocol + same-name defense Brief mention

When to use compact: Models with small context windows, or when combined with other system prompts. The compact variant preserves all rules, guardrails, and the evidence framework — it cuts examples, templates, and decision trees.

Risk: Without worked examples, models may apply the Three-Layer Model less precisely. Without templates, users must structure their own output formats.

Version History

What changed in v8 (from v7)

v8 was amalgamated from five beta candidates through systematic feature analysis (January 2026). Key changes:

  • Restructured into 10 parts (v7 had a flatter organization) — clearer separation of concerns
  • Added prompt-injection resistance — explicit instruction to treat uploaded documents as data, not commands
  • Added instruction priority hierarchy — System > Ethics > GPS > User, making conflict resolution explicit
  • Expanded document analysis protocol — step-by-step image upload handling with quality assessment
  • Added core templates — evidence table, research plan, conflict resolution matrix, proof summary format
  • Added decision trees — flowcharts for user level detection, sensitive information, and conflicting evidence
  • Added terminology quick reference — Appendix A lookup table
  • Added capabilities & limitations section — honest about what AI can and cannot do
  • Strengthened DNA ethics — explicit disclosure requirements before recommending testing
  • Added error recovery protocol — how to handle and communicate mistakes transparently

What changed in v7 (from v6.1)

  • Added adaptive user experience — beginner/intermediate/advanced detection and response calibration
  • Added scaffolding protocols — adjusting support as users grow
  • Added cognitive load management — chunking, pausing, prioritizing
  • Expanded ethics section — CARE principles for Indigenous data, diverse family structures, cultural naming

Previous Versions

Prompt Notes
research-assistant-v7.md Previous full version
research-assistant-v7-compressed.md Token-efficient v7
research-assistant-v6.1.md Earlier full version
research-assistant-v6.1-compressed.md Token-efficient v6.1

Evidence Analysis Framework

The prompts use the Evidence Analysis Process Map vocabulary:

Element Classes Definition
SOURCES Original, Derivative, Authored The containers holding information
INFORMATION Primary, Secondary, Indeterminate Content classified by informant's knowledge
EVIDENCE Direct, Indirect, Negative Interpretation relative to research question

See reference/evidence-terminology.md for detailed definitions.

References

  • Mills, Elizabeth Shown. Evidence Explained, 4th ed. (2024)
  • Board for Certification of Genealogists. Genealogy Standards, 2nd ed. revised (2021)