Skip to main content
Author: Harisfazillah Jamel
Version: 1.1.0 License: GPLv3
Status: Core Framework Documentation

1. Introduction

The CRISP² Matrix (Context-Record-Iteration-Single-Purpose-Pedagogy) is the operational engine of the DSOM protocol. It provides a structured methodology to prevent Context Decay—the phenomenon where an AI gradually loses track of complex architectural logic over a long conversation.

2. The Five Pillars (Generic Tasks)

🧩 I. Context Awareness (The Anchor)

Before any code is written, the AI must “synchronise” with the physical state of the repository.
  • The Action: Reading .agents/brain/ artifacts (palace_registry.md, task.md, walkthrough.md).
  • The Goal: Ensure the AI’s internal “Mental Anchor” matches the Human Architect’s last saved state.
  • Human Check: Always ask: “What is our current Mental Anchor?”

📝 II. Review & Record (The Audit Trail)

Logic must be committed to text before it is committed to code.
  • The Action: Updating walkthrough.md and Palace closet.md files with the reasoning behind a change.
  • The Goal: Create a permanent spatial record of why a decision was made, which serves as the “memory” for the next AI session.
  • DBP-Standard: Rekod dahulu, kod kemudian.

🔄 III. Iteration (Atomic Git Hygiene)

Large, monolithic changes are the primary cause of architectural drift and bugs.
  • The Action: Modifying only one file at a time; one commit per sub-task.
  • The Goal: Maintain High-Availability (HA) of the codebase. If an error occurs, we can roll back a single “atomic” step.
  • Commit Format: type(scope): descriptive message (e.g., feat(entities): add validation logic).

🎯 IV. Single-purpose Prompts (The Focus)

AI models perform best when given a narrow, high-depth task rather than a broad, shallow one.
  • The Action: Focusing the prompt on one specific Clean Architecture layer (e.g., just the Entities).
  • The Goal: Prevent the AI from hallucinating dependencies between unrelated layers.

🧠 V. Pedagogical Logic (The ‘Why’)

Code without explanation is “technical debt” in a multi-agent environment.
  • The Action: Explaining the security, performance, or structural logic before providing the code block.
  • The Goal: Ensure the Human Architect and the AI are in total agreement on the underlying theory.
  • Malay: Terangkan logik (Kenapa) sebelum kod (Apa).

3. Hierarchy Mapping (CRISP-DM Integration)


4. Troubleshooting Context Decay

If the AI begins to repeat errors or ignore the CRISP Strategy:
  1. STOP the current dialogue.
  2. SYNC local files to GitHub.
  3. REANIMATE using the REANIMATION-PROMPT-TEMPLATE.md.

Generated for the DSOM Sovereign Environment. UK English & DBP-standard Malay Enforced.
Deep State of Mind (DSOM) For My AI Protocol | Harisfazillah Jamel (LinuxMalaysia) | 2026-07-04 Standard: UK English | DBP-standard Bahasa Melayu Malaysia (Piawai) | GNU General Public License v3.0