System Prompt Forensics Header

System Prompt Forensics: Governance Structures in AI Developer Assistants

Treating AI system prompts as governance constitutions.

What This Research Is

  • The Hidden Constitution: System prompts are more than instructions; they are implicit constitutions defining the fundamental laws of an agent—authority, visibility, and constraints.
  • Forensic Methodology: By capturing raw HTTP traffic, we extract and normalize these constitutions into a structured, comparable format to map governance layers.
  • Discovery of PGPs: We identified recurring Prompt Governance Primitives (PGPs)—modular building blocks of predictable agent behavior.
  • Auditable Infrastructure: Moving agent governance from "vibes" to auditable infrastructure with clear authority boundaries.

Key Artifacts

Research Paper

The formal research paper detailing the PGP taxonomy and governance theory.

Technical Report

The complete technical analysis of system prompt governance across major AI assistants.

Appendix (PGPs)

A catalog of Prompt Governance Primitives identified during the research.

Executive Brief

High-level summary for technical leaders and decision-makers.

Board Brief

Strategic overview of governance risks and architectural implications.

GitHub Repository

Access the raw data, normalization scripts, and analysis artifacts.

How to Read This Work

Engineers & Researchers

Start with the Research Paper for the taxonomy and governance theory, consult the Technical Report for the comparative analysis, and explore the PGP Appendix for modular governance patterns.

Executives & Decision-Makers

Review the Executive and Board Briefs for a summary of strategic risks and the transition from "vibes" to auditable agent infrastructure.

Methodology

Our forensic workflow moves from raw observation to structured governance theory:

  1. Prompt Forensics: Capturing raw HTTP traffic to extract hidden system-level instructions.
  2. Comparative Analysis: Normalizing heterogeneous prompts into a shared structural schema to identify stable governance regimes.
  3. Governance Primitives Abstraction: Cataloging recurring, atomic control structures (PGPs) that allocate authority and bound agent action.

Disclosure

This research was produced with significant AI assistance across the analysis and synthesis pipeline, primarily using GPT-5.2 for data analysis and report generation. Final editorial oversight and methodology development were performed by the author.

The author provides the methodology and oversight but does not claim individual analytical judgments derived through the AI-driven methodology. All outputs were reviewed and verified prior to publication.

Citation

Espinoza, R. M. (2026). System Prompt Forensics: Governance Structures in AI Developer Assistants. v1.0.0. https://system-prompts-forensics.rmax.ai

View CITATION.cff

Status

Version: 1.0.0

Release Date: January 3, 2026

Stability: Publication-grade / Archival