Skip to content

GEMINI.md

Latest commit

 

History

History
228 lines (170 loc) · 8.48 KB

GEMINI.md

File metadata and controls

228 lines (170 loc) · 8.48 KB

GEMINI.md

This file provides guidance to Gemini-based AI tools (including Antigravity) when working with code in this repository.

Repository Overview

This is the Modernize framework - a production-validated system providing protocols, specialized agents, and automation for systematic AI-assisted software modernization and development. The framework supports a complete workflow: assess → plan → execute → improve.

Key Characteristics:

  • Protocol-driven architecture where each workflow is a comprehensive protocol document
  • Multi-agent coordination system with 6 specialized agents
  • Originally created for .NET Framework migrations but universally applicable
  • Emphasizes continuous improvement through retrospective analysis

Parallel Maintenance Requirement

CRITICAL: This repository maintains two parallel implementations that MUST stay in sync:

Platform Directory Guidance File
Antigravity .agent/ GEMINI.md (this file)
Claude Code plugins/ CLAUDE.md

When modifying any workflow, rule, or protocol:

  1. Always update BOTH versions - Changes to one must be mirrored to the other
  2. Check the mapping:
    • .agent/workflows/*.md ↔ plugins/autocoder/commands/*.md
    • .agent/rules/*.md ↔ plugins/modernize/agents/*.md
  3. Verify parity after changes: Both should have identical functionality
  4. Test both platforms if possible before committing

Key parallel files:

Workflow Antigravity Claude Code
/improve-test-coverage .agent/workflows/improve-test-coverage.md plugins/autocoder/commands/improve-test-coverage.md
/fix .agent/workflows/fix.md plugins/autocoder/commands/fix.md
/fix-loop .agent/workflows/fix-loop.md plugins/autocoder/commands/fix-loop.md
/monitor-workers .agent/workflows/monitor-workers.md plugins/autocoder/commands/monitor-workers.md
/monitor-loop .agent/workflows/monitor-loop.md plugins/autocoder/commands/monitor-loop.md
/full-regression-test .agent/workflows/full-regression-test.md plugins/autocoder/commands/full-regression-test.md
/review-blocked .agent/workflows/review-blocked.md plugins/autocoder/commands/review-blocked.md
/install .agent/workflows/install.md plugins/autocoder/commands/install.md

Key parallel scripts:

Script Antigravity Claude Code
start-parallel-agents.sh .agent/scripts/start-parallel-agents.sh plugins/autocoder/scripts/start-parallel-agents.sh
join-parallel-agents.sh .agent/scripts/join-parallel-agents.sh plugins/autocoder/scripts/join-parallel-agents.sh
fetch-blocked-issues.sh .agent/scripts/fetch-blocked-issues.sh plugins/autocoder/scripts/fetch-blocked-issues.sh
add-blocking-label.sh .agent/scripts/add-blocking-label.sh plugins/autocoder/scripts/add-blocking-label.sh
approve-blocked-issue.sh .agent/scripts/approve-blocked-issue.sh plugins/autocoder/scripts/approve-blocked-issue.sh
reject-blocked-issue.sh .agent/scripts/reject-blocked-issue.sh plugins/autocoder/scripts/reject-blocked-issue.sh

Workflows

Available workflows in .agent/workflows/:

Primary Workflow Commands

# 1. Assessment - Evaluate modernization readiness
/assess
# Output: ASSESSMENT.md with viability analysis across 8 dimensions

# 2. Planning - Create detailed execution strategy
/plan
# Output: PLAN.md with phase breakdown, timeline, resource allocation

# 3. Execution - Orchestrate multi-agent modernization
/modernize
# Coordinates 6 specialist agents through 7 phases

Continuous Improvement Commands

# 4. Retrospective - Analyze project history for improvements
/retro
# Output: IMPROVEMENTS.md with 3-5 evidence-based recommendations

# 5. Apply Improvements - Implement retrospective findings
/retro-apply
# Updates workflows, protocols, and automation based on IMPROVEMENTS.md

Autocoder Workflows

# Autonomous issue resolution
/fix

# Continuous issue resolution loop
/fix-loop

# Test coverage improvement
/improve-test-coverage

# Full regression testing
/full-regression-test

Recommended Workflow: /assess → /plan → /modernize → /retro → /retro-apply

Architecture

Directory Structure

.agent/
├── README.md                     # Configuration overview
├── protocols/                    # Agent behavior protocols
│   ├── adr-lifecycle.md         # ADR lifecycle (7 stages)
│   ├── testing-protocol.md      # 6-phase testing protocol
│   └── ...                      # Other protocols
├── rules/                        # Agent-specific rules
│   ├── architect.md             # Architect agent rules
│   ├── coder.md                 # Coder agent rules
│   ├── tester.md                # Tester agent rules
│   └── ...                      # Other agent rules
├── workflows/                    # Executable workflows
│   ├── fix.md            # Autonomous issue resolution
│   ├── improve-test-coverage.md # Coverage improvement
│   └── ...                      # Other workflows
└── scripts/
    └── append-to-history.sh     # Log to HISTORY.md

Multi-Agent Coordination System

6 Specialized Agents work together through protocol-defined coordination:

  1. Migration Coordinator - Strategic oversight, agent coordination
  2. Security Agent - CVE scanning, vulnerability remediation
  3. Architect Agent - Technology decisions, ADR creation
  4. Coder Agent - Implementation and migration
  5. Tester Agent - 6-phase testing protocol
  6. Documentation Agent - HISTORY.md logging, guides

Key Quality Gates

  • Security: Score ≥45/100 before migration starts
  • Build: 100% success before next stage
  • Tests: 100% pass rate before proceeding
  • Issues: All P0/P1 resolved before release

Test Coverage Report

This repository uses a persistent test-coverage.md report for efficient coverage tracking:

  • Location: test-coverage.md in project root
  • Purpose: Track coverage by functional area without re-running full analysis
  • Update: After adding tests, update the relevant section
  • Refresh: Use /improve-test-coverage --refresh to regenerate

Important Protocols

Testing Protocol (6 Phases)

All code changes must pass:

  1. Unit Tests
  2. Integration Tests
  3. Build Tests
  4. Smoke Tests
  5. Security Tests
  6. Performance Tests

Location: .agent/protocols/testing-protocol.md

Agent Logging Protocol

All agent work is logged to HISTORY.md using:

./.agent/scripts/append-to-history.sh "<title>" "<details>" "<context>" "<impact>"

Parallel Agent System

Setup and Coordination

Antigravity supports parallel agent coordination through git worktrees and workspace management:

Start parallel agents:

bash .agent/scripts/start-parallel-agents.sh [num_agents]
# Creates worktrees, generates config, attempts to launch Antigravity

Join existing session:

bash .agent/scripts/join-parallel-agents.sh
# Reads config, shows status, attempts to reopen workspaces

How it works:

  • Creates N-1 git worktrees (one coordinator stays in main repo)
  • Detects if .agent/ is a symlink and replicates to each worktree
  • Generates .agent/parallel-workspaces.json with workspace config
  • Attempts to launch Antigravity and open workspaces
  • Agents coordinate through GitHub issues and labels (working, needs-approval, P0-P3, etc.)

Key difference from Claude Code:

  • No CLAUDE_CODE_TASK_LIST_ID environment variable
  • Coordination is purely through GitHub API (labels, comments)
  • No tmux - uses Antigravity GUI workspace model

Installation Into Projects

Use /install workflow to install framework into target projects:

Option 1: Symlink (Recommended)

  • Stays updated with framework improvements
  • git pull in framework repo updates all linked projects
  • Symlinks automatically replicated to worktrees by start-parallel-agents.sh

Option 2: Copy (Custom)

  • Independent snapshot for customization
  • No automatic updates
  • Full control over protocols/rules

References