# Agent Registry & Control Verification Plan

This document outlines the steps to verify the new "Agent Registry", "Tasking", and "Portable Prompt" features in the Showcase UI.

## 1. Static Showcase (GitHub Pages / Local)

1.  Open `showcase/agents.html` in a web browser.
2.  **Registry Grid:**
    *   Verify that a grid of agents (approx 75+) is displayed.
    *   Check for filters (ALL, SPECIALIST, META-AGENT, etc.) working correctly.
    *   Test the search bar (e.g., type "Risk").
3.  **Logic Plugin:**
    *   Click the **LOGIC** button on any agent card.
    *   Verify a modal opens showing the agent's system prompt (or a placeholder if none defined).
    *   Click `[INJECT FROM LIBRARY]` to test the prompt injection simulation.
    *   Click `Update Runtime` to save changes (simulated).
4.  **Tasking & Terminal:**
    *   Click the **DEPLOY** button on any agent card.
    *   Enter a query (e.g., "Analyze AAPL") in the modal.
    *   Click **INITIATE**.
    *   Verify the **System Console** (Terminal) slides up from the bottom.
    *   Watch the simulated execution logs.
5.  **Human-in-the-Loop (Simulation):**
    *   Repeat the Tasking step multiple times or select "Human-in-the-Loop" mode if available.
    *   Wait for the "HUMAN APPROVAL REQUIRED" prompt in the terminal.
    *   Click **APPROVE** or **DENY** and verify the logs reflect the choice.

## 2. Data Consistency

1.  The data is sourced from `showcase/js/mock_data.js`.
2.  This file was generated by scanning the `core/agents/` directory.
3.  Verify that `Meta_Orchestrator`, `Risk_Assessment_Agent`, and new agents like `QuantumScenarioAgent` are present.

## 3. Fallback Mechanism

1.  The system automatically detects if the backend API is unavailable (which is true for static pages).
2.  It falls back to `window.MOCK_DATA` instantly.
3.  No 404 errors should appear for the main data load.