Claude to Gemini Context Handoff

Overview

  • Project: Aloysius Productions
  • Status: Active — transferring organizational principles across AI agents
  • Goal: Apply workflow and principles regardless of which agent is in use

The “Core + Adapter” Architecture

The most robust approach separates immutable organizational rules from agent-specific mechanics:

Core Layer (Agent-Agnostic)

CORE_PRINCIPLES.md — The organization’s “Constitution”:

  • Coding style and architectural patterns
  • Dogfooding protocols
  • Compliance requirements
  • Quality gates

Adapter Layer (Agent-Specific)

FileAgentAdditions
.claude.mdClaude CodeIncludes CORE_PRINCIPLES + Claude-specific instructions
.gemini.mdGemini CLIIncludes CORE_PRINCIPLES + Gemini-specific instructions (e.g., leverage 2M context window)
# Master file
_instructions.md
 
# Symlinks for each agent
ln -s _instructions.md .gemini.md
ln -s _instructions.md .claude.md

Benefit: Update standards in one place, every agent gets the update instantly. Eliminates version drift.

Gap-Filling Prompts

Prompt A: Translation Audit

Identify Claude-specific references (Artifacts, tool limitations) that confuse Gemini. Expand compressed shorthand where Gemini’s larger context allows.

Prompt B: Principle Stress-Test

Act as “Compliance Officer” — find where principles are stated but not operationalized (e.g., “TDD” mentioned but no explicit “write tests before code” instruction).

Prompt C: Context Expansion

Leverage Gemini’s 1M+ token context — scan entire project structure and recommend which docs to permanently include for “God-mode” understanding.

Key Differences

FeatureClaude WorkflowGemini Workflow
Context strategySelective/condensed (save tokens)Comprehensive/full repo (feed the architecture)
File management.claude.md specific rulesSymlinked CORE_PRINCIPLES across agents
Gap fillingManual promptingRecursive self-analysis