Regen AI Claude Configuration: Architecture & New Skills

Regen AI
1

Regen AI Claude Configuration: Architecture & New Skills

Overview

As part of the Regen AI development (see Announcing Regen AI and
Regen KOI MCP: The Knowledge Brain of Regeneration), we’ve established a structured configuration system for Claude Code that enables
tiered access, reusable skills, and domain-specific context injection.

This post documents the architecture and introduces two new skills now available to the core team.

Configuration Repository Architecture

We maintain two complementary repositories:

Public Configuration: GitHub - regen-network/regen-ai-claude: Regen AI marketplace for claude code plugins.

regen-ai-claude/
├── plugins/ # MCP server plugins
│ ├── koi/ # KOI Knowledge Commons
│ ├── ledger/ # Regen Ledger queries
│ └── registry-review/ # Registry automation
├── public/
│ ├── CLAUDE.md # Public tier context
│ └── .mcp.json # Basic MCP config
└── README.md

Purpose: Publicly accessible Claude Code plugins and basic configuration. Anyone can use these to connect to Regen AI infrastructure.

Core Configuration: https://github.com/regen-network/regen-ai-core

regen-ai-core/
├── CLAUDE.md # Core team extended context
├── contexts/
│ ├── ECOCREDIT.md # Ecocredit module knowledge
│ ├── GOVERNANCE.md # Governance patterns
│ ├── KOI_KNOWLEDGE.md # KOI protocol details
│ └── REGEN_LEDGER.md # Ledger architecture
├── skills/
│ ├── code-review/ # Code review automation
│ ├── credit-analysis/ # Credit class analysis
│ ├── ledger-query/ # Ledger query patterns
│ ├── weekly-digest/ # Weekly digest generation
│ ├── living-language-tone/ # NEW: Executive communication
│ └── metaprompt/ # NEW: Prompt engineering
└── agents/
└── templates/ # Agent configurations

Purpose: Extended context and specialized skills for core team members. Accessible via @regen.network authentication.

New Skills Added

  1. Living Language Tone Skill

Location: https://github.com/regen-network/regen-ai-core/blob/main/skills/living-language-tone/SKILL.md

Purpose: Transform drafts into credible, executive-ready language that carries living systems logic without esoteric terminology.

When to Use:

  • Investor decks and board materials
  • Partner communications
  • Governance documentation
  • Technical specs for non-technical audiences

Core Transformation Pattern:
┌────────────────┬──────────────────────────────────────────────────────────┐
│ From │ To │
├────────────────┼──────────────────────────────────────────────────────────┤
│ “optimize” │ fit, durability, adaptability, resilience │
├────────────────┼──────────────────────────────────────────────────────────┤
│ “tokenomics” │ incentive design, coordination infrastructure │
├────────────────┼──────────────────────────────────────────────────────────┤
│ “governance” │ decision-making infrastructure, legitimacy mechanisms │
├────────────────┼──────────────────────────────────────────────────────────┤
│ “regenerative” │ long-term stewardship, systems health, adaptive capacity │
└────────────────┴──────────────────────────────────────────────────────────┘
Example:

Before: “The $REGEN token optimizes ecosystem value capture and aligns stakeholder incentives through governance participation.”

After: “The $REGEN token coordinates decision-making across network participants and directs transaction fees toward activities that strengthen the network’s
long-term capacity to verify ecological outcomes.”

  1. Meta-Prompt Engineering Skill

Location: https://github.com/regen-network/regen-ai-core/blob/main/skills/metaprompt/SKILL.md

Purpose: Systematically improve AI output quality through structured prompt design, context engineering, and multi-phase analysis.

When to Use:

  • Refining vague requests into precise directives
  • Extracting structured data from unstructured sources
  • Designing complex multi-tool workflows
  • Creating reusable prompt templates

The Meta-Prompt Anatomy:

  1. Context Block — Domain, input source, constraints
  2. Objective Block — Primary output, success criteria
  3. Extraction Framework — Categories, hierarchy, priorities
  4. Output Structure — Format, length, citations
  5. Quality Gates — Verification checklist
  6. Meta Instructions — Edge cases, uncertainty handling

Extraction Primitives:

  • EXTRACT-TEMPORAL — Past/Present/Future analysis
  • EXTRACT-STAKEHOLDER — Earth/Community/Customers/Investors lens
  • EXTRACT-DECISION — Options/Trade-offs/Recommendation structure

How It Connects to HTTP Config Endpoint

These repositories feed into the feat: Implement Claude Config HTTP endpoint by glandua · Pull Request #3 · DarrenZal/koi-research · GitHub (now merged), which enables:

┌─────────────────────────────────────────────────────────────┐
│ Claude Code Session (any platform) │
│ GET /api/koi/claude-config/bundle │
└─────────────────────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────────────────────┐
│ Tier Determination │
│ - Unauthenticated → public tier │
│ - @regen.network email → core tier │
└─────────────────────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────────────────────┐
│ Response: Merged CLAUDE.md + contexts + skills │
└─────────────────────────────────────────────────────────────┘

This means core team members get automatic access to these skills and contexts in any Claude Code session—local, cloud, or co-work—without manual configuration.

Using the Skills

In Claude Code

Skills are invoked contextually. When working on investor materials:

“Apply the living-language-tone skill to this draft…”

When designing a complex extraction workflow:

“Use the metaprompt skill to structure this analysis…”

Contributing New Skills

Skills follow a standard format:

[Skill Name]

Purpose

What this skill does

When to Use

Trigger conditions

Methodology

Step-by-step process

Examples

Before/after demonstrations

Quality Checklist

Verification criteria

Submit new skills via PR to https://github.com/regen-network/regen-ai-core.

Related Resources

Discussion

  • What skills would be most useful for your work?
  • Are there communication patterns or analysis workflows we should systematize?
  • Feedback on the living-language-tone transformations?

Configuration architecture developed as part of Regen AI Phase 2 — February 2026

1 Like