Prime-Rule-Workflow-for-AI-Agent-RnD-Boost
Health Pass
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Community trust — 37 GitHub stars
Code Warn
- Code scan incomplete — No supported source files were scanned during light audit
Permissions Pass
- Permissions — No dangerous permissions requested
No AI report is available for this listing yet.
Prime Protocols: global rules, workflows, skills and hooks that make AI agents (Claude Code, Antigravity and others) work with rigour on research and engineering.
title: "Convergent Cognitive Architecture (V5)"
abstract: "Project landing page and architecture overview of the Prime Protocols V5 for agentic engineering systems."
date: 2026-06-23
keywords: ["agentic", "cognitive", "architecture", "rules", "workflows"]
status: "active"
🌌 Convergent Cognitive Architecture (V5)
The Prime Protocols V5: A dual-layered agentic AI coding engine combining a structured Mental OS (Global Rules) and a monotonic Execution Playbook (Workflow Engine).
This repository contains the complete specification and phase-by-phase execution manuals for running high-precision engineering agents (such as Antigravity, Claude Code and other AI agents) across complex, multi-domain programming environments.
Version note: the files here are V5, written for Antigravity with Gemini. V6 is in preparation. It is built from one source for any AI agent host, starting with Claude Code and Antigravity, and adds skills, hooks and a one-command installer. It will be published here as a release.
🚀 Architecture Overview
The system is split into two co-dependent layers:
- The Global Rules (Mental OS): Governing identity, task decomposition, tri-layer awareness, safety constraints, triad debates, and data preservation (Delta Integrity).
- The Workflow Engine: A convergent state machine guiding the agent through deterministic steps to guarantee that clarity and verification increase monotonically with each phase.
flowchart TD
subgraph HumanDomain["Human Domain (Inputs & Commands)"]
UserInput["New Task / Info"]
UserCommand["Trigger /crystallize"]
end
subgraph AIBrain["AI Cognitive Engine (Internal Processes)"]
P1["Phase 1: Explore"]
P2["Phase 2: Parse"]
P3["Phase 3: Plan"]
P4["Phase 4: Execute"]
P5["Phase 5: Verify"]
P6["Phase 6: Align"]
Debate{"Triad Debate"}
end
subgraph AIOutput["AI Output (Visible to Human)"]
Halt["Gated Halting (Ask User for Help)"]
P7["Phase 7: Crystallize (Write Heuristics)"]
Complete["Mission Complete (Await Next Task)"]
end
UserInput -->|Initiates Mission| P1
P1 -->|Context Known| P2
P2 -->|Task Graph Mapped| P3
P3 -->|Aligned Strategy| P4
P4 -->|Delta Matrix Applied| P5
P5 -->|All Subtasks Verified| P6
P6 -->|Aligned| Complete
Complete -.->|Optional Command| UserCommand
UserCommand -->|Extracts Wisdom| P7
P1 -.->|Context Gaps| Debate
P2 -.->|Unknown Origin| Debate
P5 -.->|Loop Count >= 3| Debate
Debate ==>|Ruling Issued| P3
P5 -.->|Diagnostic Failure / Subtask Incomplete| P3
P6 -.->|Architectural Drift| P3
Debate ===>|Unresolved| Halt
P5 ===>|Blocked / Max Loops Reached| Halt
P6 ===>|Alignment Failed| Halt
Halt -.->|User Answers| UserInput
📂 Repository Structure
- GEMINI_v5.md: The core system prompt containing the 8 Laws of Tactical Execution, Multi-Task Orchestration protocols, and the primary state topologies.
- Workflows_for_V5/: Detailed playbook guides defining agent behavior in the automatic state machine loop:
- explore.md (Phase 1: Context Acquisition)
- parse.md (Phase 2: Task Decomposition)
- plan.md (Phase 3: State Tracker & DoD)
- execute.md (Phase 4: Delta Matrix Rules)
- verify.md (Phase 5: Gated Loops)
- align.md (Phase 6: Global Alignment)
- debate.md (Interstitial Triad Resolution)
- Independent_Workflows/: Playbooks for workflows triggered manually (outside the automatic state machine loop):
- crystallize.md (Phase 7: Universal Heuristics extraction)
- consolidate.md (Project Knowledge Consolidation)
- .ai/: Persistent workspace configuration and session logs folder:
- local-log.md: Record of design decisions, milestones, and session logs.
- local-rules.md: Strategic local guidelines and empirical heuristics derived from post-execution crystallization.
- references.md: Trajectory records, references, and failure/variance analyses.
🛠️ How to Import into Agent IDEs
To configure your agent (e.g., Cursor, Claude Code, or Antigravity IDE) to use this system:
- System Prompt / Global Rules: Copy the contents of GEMINI_v5.md into your IDE's system rules setting or save it as
.cursorrules/.windsurfrulesat your project root. - Workflows: Place the
Workflows_for_V5directory under.gemini/config/global_workflows/or reference them locally to guide phase transitions. - Run Commands:
- Automatic Loop:
/explore,/parse,/plan,/execute,/verify,/align,/debate - Manual / Independent:
/crystallize,/consolidate
- Automatic Loop:
👤 Author
- kv3100123 - @kv3100123
⚖️ License
This project is open-sourced under the MIT License. See LICENSE if applicable.
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found