Airsc is free for everyone. If it helps you, support it with $2–$20.

Support
VISign in
PluginCautionAgent workflow & memorySome setup

LLM-Autonomous-Agent-Plugin-for-Claude

bejranonda · bejranonda/LLM-Autonomous-Agent-Plugin-for-Claude

This Claude Code plugin provides an autonomous, self-learning agent system with 36 specialized subagents and a 4-tier architecture for analysis, planning, execution, and validation of code, focusing on privacy and security.

Autonomous self-learning agent plugin for Claude Code — 36 specialized subagents in a 4-tier architecture (analysis, planning, execution, validation), 27 skills, 41 slash commands across 10 categories. Pattern learning, real-time dashboard, full-stack validation, OWASP security. 100% local processing, privacy-first, cross-platform.

28 stars15License: No licenseUpdated 4 months ago0 opens

Preview

▍

What you can do with it

  • Automatically implement features from high-level requirements.
  • Conduct full-stack validation and OWASP security checks on codebases.
  • Continuously improve code quality and development processes through pattern learning.

Safety check

Caution

Usable, but read the notes first.

  • ●No clear license: check with the author before using it in a commercial product.

Automated scan of the repo files. Always review code you run.

From the README

🚀 Autonomous Agent for Claude Code v8.4.5

What if your AI agent got smarter and faster with every task?

🚀 Installation • 🎯 Quick Start • 📚 Commands • 🛠️ Features • 📊 Dashboard

Automatic Learning System | Privacy-First | Production-Ready Analysis | Real-Time Monitoring | KPI Intelligence | OWASP Security | Auto Fixes

  • Every Task Makes It Smarter
  • No configuration required.
  • No manual training.
  • Just automatic continuous improvement across ALL models.
  • 60-70% cost reduction with comprehensive KPI tracking 🆕

The autonomous agent is now smarter and more organized than ever, with revolutionary category-based commands that learn from every task and comprehensive metrics intelligence that tracks every optimization! 🚀

Autonomous Agent Dashboard check-in! 🤖 7 Active Agents, 5 Skills, and a 97.5/100 Quality Score.

Complete Workflow Example:

# What do you need
/dev:auto "add user authentication"    # Implement
# Prove and Release the project
/dev:release --minor                   # Release (2-3 min)
# Done! From requirement to released

🌟 What Makes Revolutionary?

Experience enterprise-grade autonomous intelligence that continuously evolves with every task.

🧠 Core Innovation: Revolutionary Four-Tier Architecture

A paradigm shift from static tools to living intelligence with 36 specialized agents across 4 collaborative groups (including the master orchestrator):

🏗️ Four-Tier Group Architecture (v8.0.0+)

  • Group 1 - Strategic Analysis (Brain): 8 agents analyze and recommend with confidence scores
  • Group 2 - Decision Making (Council): 2 agents evaluate and create optimal execution plans
  • Group 3 - Execution (Hand): 14 agents implement with comprehensive metrics
  • Group 4 - Validation (Guardian): 7 agents optimize and provide continuous feedback

🚀 Revolutionary Breakthroughs

🎯 Enterprise-Grade Autonomous Operation

  • 98% Success Rate: Complete independence with zero human intervention
  • Intelligent Coordination: Seamless agent collaboration across all 4 groups
  • Inter-Group Learning: Automatic knowledge transfer and continuous improvement

🧠 Advanced Learning System (95/100 Quality Score)

  • Exponential Learning: 35% improvement per 10 similar tasks (133% faster)
  • Pattern Recognition: 94% accuracy in identifying successful approaches
  • Predictive Skill Selection: 92% accuracy in optimal skill combinations

📈 Comprehensive KPI Intelligence (v7.5.0)

  • 11 KPIs Across 5 Categories: Performance, Cost, Quality, User Experience, System Health
  • Unified Dashboard System: Single comprehensive interface consolidating all monitoring views
  • 5 Tabbed Sections: Overview, Analytics, Token Optimization, KPI & Reports, System Health
  • Mobile-Responsive Design: Full functionality on all devices with real-time updates
  • Export Capabilities: JSON, CSV, and PDF report generation
  • 60-70% Cost Reduction: Automatic token optimization with ML-based strategies

🛡️ Full-Stack Auto-Fix Intelligence

  • 80-90% Auto-Fix Success: 24 patterns automatically resolve common issues
  • Multi-Language Mastery: 40+ linters across 15+ programming languages
  • OWASP Top 10 Security: Complete vulnerability coverage with automated remediation

📊 Revolutionary Advantages Matrix

CapabilityTraditional ToolsAutonomous Agent v7.5.0
IntelligenceStatic analysis✅ Living AI that evolves
AutonomySemi-automated✅ Complete independence
LearningFixed patterns✅ Exponential improvement
CoordinationSingle tools✅ 36 agent ecosystem (4 groups)
AnalyticsBasic metrics✅ Comprehensive KPI intelligence
ValidationManual checks✅ 80-90% auto-fix + 5-layer validation
PrivacyCloud processing✅ 100% local processing
PerformanceHours of analysis✅ Seconds with insights
Cost OptimizationManual optimization✅ 60-70% automatic reduction

🏆 Key Achievements

  • ✅ Four-Tier Architecture - Revolutionary separation of analysis, decision, execution, validation
  • ✅ Production Certification - 100/100 validation score with zero blockers
  • ✅ Unified Storage Revolution - 90% performance boost, eliminated 47+ scattered files
  • ✅ Pattern-Based Intelligence - 30+ learned patterns driving optimal decisions
  • ✅ Real-Time Monitoring - Interactive dashboards with 30-second auto-refresh

📊 Quantified Impact

  • Analysis Speed: 5-15 minutes → 5-15 seconds (40x faster)
  • Learning Accuracy: 70% → 92.3% (22% improvement)
  • Auto-Fix Rate: 20% → 80-90% (4x improvement)

🌟 What is New?

EVOLUTION OF EXCELLENCE: From Basic Analysis to Enterprise-Grade Autonomous Intelligence

v8.4.5 - Quality Metrics Correctness

  • Fixed quality_control_check.py reporting Successful Imports: 1 (out of ~82) due to a sys.path-sensitive importlib.import_module call. Replaced with file-based importlib.util.spec_from_file_location, so imports resolve by absolute path regardless of how the script is launched. Metric now correctly reports 82; Overall Score: 100/100 unchanged (score uses ast.parse, not imports).
  • Same file's command count was off by the v8.4.1 non-recursive-glob bug (*.md missed the 10 category subfolders) - now reports 41 commands (was 1). Documentation coverage formula also stopped counting archived reports under data/reports/archive/ against the active-docs percentage.
  • pattern_storage.py _ensure_directory() no longer writes a bare [] that pre-emp
View source on GitHub