intelligent-routing

Files (1)

• SKILL.md

SKILL.md (first lines)

# Intelligent Agent Routing (With Self-Evolving Soul 🦅)

**Purpose**: Automatically analyze user requests, retrieve past "Soul Memory", and route them to the most appropriate specialist agent(s).

## Core Principle

> **The AI acts as an evolving entity**, retrieving wisdom from the **ReasoningBank** before every specialist invocation.

## How It Works

### 1. Request Analysis & Soul Retrieval

Before responding to ANY user request, perform the following:

```mermaid
graph TD
    A[User Request] --> B[Phase -1: SOUL RETRIEVAL]
    B --> RB[(ReasoningBank)]
    RB -->|Top 3 Patterns| C[ANALYZE & ROUTE]
    C --> D[Keywords + Domains]
    D --> E[SELECT AGENT]
    E --> F[Inject Soul Memory into Agent Context]
    F --> G[GO!]
```

### 2. Agent Selection Matrix

**Use this matrix to automatically select agents:**

| User Intent         | Keywords                                   | Selected Agent(s)                           | Auto-invoke? |
| ------------------- | ------------------------------------------ | ------------------------------------------- | ------------ |
| **Authentication**  | "login", "auth", "signup", "password"      | `security-auditor` + `backend-specialist`   | ✅ YES       |
| **UI Component**    | "button", "card", "layout", "style"        | `frontend-specialist`                       | ✅ YES       |
| **Mobile UI**       | "screen", "navigation", "touch", "gesture" | `mobile-developer`                          | ✅ YES       |
| **API Endpoint**    | "endpoint", "route", "API", "POST", "GET"  | `backend-specialist`                        | ✅ YES       |
| **Database**        | "schema", "migration", "query", "table"    | `database-architect` + `backend-specialist` | ✅ YES       |
| **Bug Fix**         | "error", "bug", "not working", "broken"    | `debugger`                                  | ✅ YES       |
| **Test**            | "test", "coverage", "unit", "e2e"          | `test-engineer`                             | ✅ YES       |
| **Deployment**      | "deploy", "production", "CI/CD", "docker"  | `devops-engineer`                           | ✅ YES       |
| **Security Review** | "security", "vulnerability", "exploit"     | `security-auditor` + `penetration-tester`   | ✅ YES       |
| **Performance**     | "slow", "optimize", "performance", "speed" | `performance-optimizer`                     | ✅ YES       |
| **Product Def**     | "requirements", "user story", "backlog", "MVP" | `product-owner`                             | ✅ YES       |
| **New Feature**     | "build", "create", "implement", "new app"  | `orchestrator` → multi-agent                | ⚠️ ASK FIRST |
| **Complex Task**    | Multiple domains detected                  | `orchestrator` → multi-agent                | ⚠️ ASK FIRST |

### 3. Automatic Routing Protocol

## TIER 0 - Automatic Analysis (ALWAYS ACTIVE)

Before responding to ANY request:

```javascript
// Pseudo-code for decision tree
async function analyzeRequest(userMessage) {
    // 1. Phase -1: Soul Retrieval (NEW)
    const soulMemory = await retrieveSoulMemory(userMessage, 3);

    // 2. Classify request type
    const requestType = classifyRequest(userMessage);

    // 3. Detect domains
    const domains = detectDomains(userMessage);

    // 4. Determine complexity
    const complexity = assessComplexity(domains);

    // 5. Select agent(s)
    let selectedAgents;
    if (complexity === "SIMPLE" && domains.length === 1) {
        selectedAgents = selectSingleAgent(domains[0]);
    } else if (complexity === "MODERATE" && domains.length <= 2) {
        selectedAgents = selectMultipleAgents(domains);
    } else {
        selectedAgents = ["orchestrator"];
    }

    // 6. Prepend Soul Memory to first prompt (NEW)
    return {
        agents: selectedAgents,
        promptInjection: soulMemory
    };
}
```

### 4. The Handshake Routing Protocol (Efficiency Model 🦅)

**Goal**: Maximize accuracy while minimizing token cost using tiered model coordination.

| Tier | Agent | Model | Task | Autonomy |
| :--- | :--- | :--- | :--- | :--- |
| **Tier 1: SCOUT** | `explorer-agent` | `lite` | Discovery, Search, List Files | **FULL** (Autonomous) |
| **Tier 2: BUILDER**| Specialist Agent | `pro` | Implementation, Reasoning | **GUIDED** |
| **Tier 3: SENTRY** | `test-engineer` | `flash` | Verification, Linting | **STRICT** |

#### The Implementation Flow:
1. **Scout Discovery**: If the task is unclear or spans multiple files, the Scout first runs `find_by_name` or `grep_search`.
2. **Context Trimming**:
    - **General Tasks**: 50-line window.
    - **Architectural Tasks**: 100-line window around matches.
3. **The Handshake**: The Scout passes the *trimmed* context to the Builder.
4. **Sentry Check**: If the Sentry (cheap model) detects a regression in the Builder's work: **STOP & ASK**.

---

## 5. Scout Autonomy Rules

1. **Scout is authorized to**:
    - List any directory.
    - Search for patterns in codebase.
    - Read file outlines.
2. **Scout is FORBIDDEN to**:
    - Edit files.
    - Access external URLs.
    - Commit code.

---

## 6. Response Format

From ashrafmusa/agenticana, MIT.

More from ashrafmusa/agenticana