ECC - Open-Source Cross-Platform AI Agent Performance Optimization System

What is ECC

ECC (Everything Claude Code) is a cross-platform AI Agent performance optimization system. The system provides Claude Code、Codex、Cursor and 7+ other AI programming tools with 63 professional Agents, 249 on-demand loading Skills, cross-session memory persistence, and the AgentShield security audit tool, enabling automated workflows through the Hook mechanism, transforming the AI programming assistant into an AI team with clear division of labor.

ECC's Main Features

• 63 Professional Agents: Covering roles across the entire development chain, including architecture design, code review, security audit, test-driven development, build error fixing, and documentation updates.
• 249 On-Demand Loading Skills: Dynamically match skills based on the project's tech stack. TypeScript projects automatically load the TS review Skill, Python testing scenarios trigger the TDD Skill, avoiding context overflow.
• Cross-Session Memory Persistence: Automatically store and retrieve context between different sessions through the Hook mechanism, enabling uninterrupted continuous workflows.
• AgentShield Security Audit: Built-in 1282 tests and 102 static analysis rules, scanning for credential leaks, misconfigurations, and injection risks at millisecond speed.
• Continuous Learning v2: An intuition-based learning system that automatically extracts reusable patterns from sessions, generates Instincts, and clusters them into Skills.
• Verification Loops and Checkpoints: Supports /checkpoint saving verification states,/quality-gate quality gates, and the /eval evaluation framework.
• Multi-Agent Collaboration: Supports /multi-plan、/multi-execute and other multi-model collaborative workflow commands.
• Dashboard GUI: Provides a Tkinter desktop application supporting theme switching, font customization, and visual browsing of components.

ECC's Technical Principles

• On-Demand Loading Architecture: Uses a Manifest-driven installation pipeline, with state storage tracking installed components, injecting Skills into context only when the corresponding tech stack is detected.
• Hook Event System: Based on the Hook mechanism of Claude Code v2.1+, triggering automated scripts at 8 event points including SessionStart, PreToolUse, PostToolUse, and Stop.
• Context Budget Management: Controls the session startup context limit through environment variables such as ECC_SESSION_START_MAX_CHARS , combined with strategic compression suggestions to prevent Token overflow.
• AgentShield Adversarial Scanning：--opus mode launches a three-agent adversarial pipeline consisting of a red team Agent (finds vulnerabilities), a blue team Agent (fixes them), and an auditor Agent (summarizes and evaluates).
• Selective Installation Architecture: Supports minimal、core、full and other multi-tier installation configurations, allowing users to copy rules, Agents, and Skills as needed.
• Cross-Harness DRY Adapter: Hook events from platforms such as Cursor are converted to Claude Code format through adapter.js , reusing the same set of script logic.

How to Use ECC

• Plugin Installation：
  • In Claude Code, execute /plugin marketplace add https://github.com/affaan-m/ECC to add the marketplace source.
  • Execute /plugin install ecc@ecc to install the plugin, automatically loading Skills, commands, and Hooks.
  • Manually copy the required rules/ directory to ~/.claude/rules/ecc/(the plugin does not support automatic distribution of rules).
• Manual installation：
  • Clone the repository git clone https://github.com/affaan-m/ECC.git。
  • Copy as needed agents/、skills/、commands/ to ~/.claude/ the corresponding directory.
  • Run ./install.sh --profile minimal --target claude to install the Hook runtime.

ECC's core advantages

• Broadest cross-platform coverage: Native support for 7+ Harnesses including Claude Code, Codex, Cursor, OpenCod, etc., with one configuration reused across multiple tools.
• Intelligent context management: 249 Skills loaded on demand, combined with MCP tool count control recommendations (no more than 10 MCPs per project), effectively protecting the 200K context window.
• Production-grade security built in: AgentShield provides static analysis with 98% coverage, supporting --fix automatic fixes and --opus in-depth adversarial auditing.
• Self-evolution capability: The continuous learning system extracts patterns from real development sessions, making the configuration increasingly tailored to personal workflows with use.

• Project official website：https://ecc.tools/
• GitHub repository：https://github.com/affaan-m/ECC

Comparison of ECC with similar competitors

ECC's application scenarios

• Full-stack project development: A complete AI-assisted development pipeline from requirements planning and architecture design to TDD coding, code review, and E2E testing.
• Multi-technology-stack teams: Simultaneously maintain multilingual projects such as TypeScript, Python, Go, Java, Rust, Swift, and PHP, automatically matching the corresponding review Agent.
• Secure and compliant development: In financial and enterprise-grade projects, use AgentShield to prevent key leakage, configuration errors, and injection attacks.
• Long-term maintenance: Leverage cross-session memory and continuous learning to keep AI's understanding of large codebases across multiple sessions.
• Hackathons and rapid prototyping: Drawing on the author's championship-winning experience, complete the entire chain from customer research to product prototype within a few hours.