# 🏗️ Skill & Tool Tracker Architecture & Technical Specifications

## 1. System Overview
Skill & Tool Tracker is a high-performance, local-first developer platform combining **AI Agent Skills Observability**, a **Dynamic Tool Calling & DAG Engine**, and a **Model Context Protocol (MCP) Gateway**.

```
+-------------------------------------------------------------------------+
|                          AI AGENT CLIENTS                               |
|   (Antigravity IDE, Claude Code, Cursor, Windsurf, Continue, Goose)     |
+------------------------------------+------------------------------------+
                                     |
                [JSON-RPC 2.0 stdio] | [HTTP / REST API]
                                     v
+-------------------------------------------------------------------------+
|                  SKILL & TOOL TRACKER PLATFORM                          |
|                                                                         |
|  +-----------------------+  +----------------------------------------+  |
|  |   MCP SERVER GATEWAY  |  |       DYNAMIC TOOL RUNTIME ENGINE      |  |
|  |  - search_skills      |  |  - JavaScript Sandbox (Node.js VM)     |  |
|  |  - get_skill_guide    |  |  - Python Sandbox                      |  |
|  |  - execute_tool       |  |  - Shell & HTTP Dispatchers            |  |
|  |  - plan_tool_dag      |  |  - DAG Dependency Resolver             |  |
|  +-----------------------+  +----------------------------------------+  |
|                                                                         |
|  +-----------------------+  +----------------------------------------+  |
|  |  LOG COLLECTOR ENGINE |  |       WEB SPA DASHBOARD (PORT 17830)   |  |
|  |  - Multi-agent Parser |  |  - Pure Vanilla ES6+                   |  |
|  |  - Dedup Window (2m)  |  |  - Realtime Live-Sync Watcher          |  |
|  |  - Conflict Detector  |  |  - Dynamic Bilingual i18n (VI / EN)    |  |
|  +-----------------------+  +----------------------------------------+  |
|                                                                         |
|  +-------------------------------------------------------------------+  |
|  |                        LOCAL DATA STORAGE                         |  |
|  |   skills/ (186 skills)  |  data/tools_registry.json (44 tools)    |  |
|  |   data/tool_executions  |  dashboard/skill_catalog.json           |  |
|  +-------------------------------------------------------------------+  |
+-------------------------------------------------------------------------+
```

---

## 2. Core Subsystems

### A. Skills Observability & Discovery (`skills/`)
- **186 Registered Skills**: Covers Software Engineering, Mobile, Cloud, Data, Bio/Science, and the full Superpowers methodology suite.
- **On-Demand RAG Discovery**: Search skills with keyword and semantic scoring to minimize context window bloat (< 200 tokens vs 30k tokens).

### B. Dynamic Tool Execution & DAG Orchestration (`tools/`)
- **44 Builtin Tools**: Formats, encoders, calculators, translators, HTTP fetchers, and validators.
- **DAG Pipeline Resolver**: Decomposes complex user goals into dependency graphs and executes nodes sequentially or in parallel.

### C. Model Context Protocol (MCP) Server (`mcp-server/`)
- Implements MCP Specification `2024-11-05` over `stdio`.
- Exposes 6 high-level meta-tools and 2 resource endpoints.

### D. Single Page Application Dashboard (`dashboard/`)
- Zero npm/webpack build dependencies (instant load).
- 6 Interactive Views: Overview Stats, Audit Logs, Skill Catalog, Risk Governance, Action Plans, Tool Matrix.
- 100% Dynamic Bilingual Synchronization (VI / EN).