Issue

You use AI to code — Cursor, Copilot, Claude, Windsurf, or any IDE. But AI often:

• ❌ Does not understand the codebase — suggests the wrong file, wrong function
• ❌ Forgets context — new session = forgets all previous decisions
• ❌ Wastes tokens — reads the entire file when only 1 function is needed
• ❌ Inconsistent code — generates a different style each time
• ❌ Loses focus — starts task A and wanders off to task B

Solution

DOMYH Awesome Code = 3 levels of infrastructure for AI agents:
🧠 Level 1 — AWF (Agent Workspace Framework) Open Source · MIT

• 103 specialized skills: TypeScript, Python, Go, React, Vue, Next.js, Docker, K8s...
• 44 automated commands: /code, /debug, /test, /deploy, /plan, /review...
• 11 personas: Developer, Architect, Auditor, Debugger, Tester...
• 27 IDE configs from 1 single set — Cursor, Claude, Copilot, Windsurf, Cline, Codex, Kiro, JetBrains...
• Decision trees for 11 skills — AI chooses the appropriate approach
• Progressive loading — only loads necessary skills, saving tokens
• Install once, use with any IDE
• Deep workflows perform best when combined with HSA MCP Server

🔍 Level 2 — HSA MCP Server Free · npm · v1.7.4

• Hybrid Search: BM25 + Vector + CodeGraph — find precise code in large projects
• Session Memory: AI remembers context, decisions, and conventions across sessions
• Drift Detection: Automatically alerts when the AI loses focus on the main task
• PageRank: Ranks files by level of importance
• Doc Fetcher: Retrieves real docs of 80+ libraries (Next.js, Prisma, React...)
• Design Intelligence: Analyzes CSS, WCAG accessibility, screenshot comparison
• Knowledge Graph: 93 semantic edges for deep semantic analysis
• Memory Guard (v1.7.4): BM25 100K term cap, hapax pruning, RSS guard — stability for large projects
• 15 MCP tools — works with any IDE that supports MCP protocol

🖥️ Level 3 — HSA Companion (VS Code Extension) Free · Open VSX

• Auto-Accept: Automatically approves safe actions, only asks when dangerous
• Telegram Remote: Control coding sessions remotely via Telegram bot
• Session Tracking: Knows when AI is done, sends push notifications
• Dashboard: Web UI for real-time session monitoring

Distribution Model

| Component | Source | Distribution | Cost |
|:-----------|:-------|:----------|:--------|
| AWF (configs, skills, workflows) | 🟢 Open Source (MIT) | GitHub + npm | Free |
| HSA (MCP context engine) | 🔒 Closed Source | npm @nockdev/hsa) | Free |
| Companion (VS Code extension) | 🔒 Closed Source | Open VSX | Free |
| nock-cli (CLI tool) | 🟢 Open Source | npm @nockdev/cli) | Free |
> 💡 AWF functions independently with any IDE (rules, skills, personas). However, deep workflows such as /ap (audit), /debug, /code are designed to leverage