SuperCharged-Claude-Code-Upgrade/manual-approval-test.md

# Approval Flow Integration Test Guide

This document provides step-by-step instructions for manually testing the approval flow fix.

## Test Environment Setup

1. Start the server:
   ```bash
   cd /home/uroma/obsidian-web-interface
   node server.js
   ```

2. Open browser to: `http://localhost:3000/claude-ide`

3. Open browser DevTools (F12) and go to Console tab

## Test Scenario 1: Basic Approval Flow

### Step 1: Initiate Command
- Type in chat: `run ping google.com`
- Press Enter

**Expected Result:**
- Message appears in chat
- Claude responds with approval request
- Approval card appears with Approve/Reject buttons

**Console Should Show:**
```
[ApprovalCard] Component loaded
```

### Step 2: Verify Approval Request
- Check that approval card shows:
  - Robot emoji (🤖)
  - Command: `ping google.com`
  - Explanation (if provided)
  - Approve button
  - Custom Instructions button (if explanation provided)
  - Reject button

**In DevTools Console, type:**
```javascript
window._pendingApprovals
```

**Expected:** Object with approval ID containing:
- `command`: "ping google.com"
- `sessionId`: current session ID
- `originalMessage`: user's message

### Step 3: Click Approve
- Click the "Approve" button

**Expected Result:**
- Approval card disappears
- System message appears: "✅ Approval sent - continuing execution..."
- "yes" message sent to Claude

**Console Should Show:**
```javascript
// Check WebSocket message in Network tab
// Should see message with:
{
  type: 'command',
  sessionId: '<session-id>',
  command: 'yes',
  metadata: {
    isApprovalResponse: true,
    approvalId: '<approval-id>',
    originalCommand: 'ping google.com'
  }
}
```

**Verify Pending Approval Cleaned:**
```javascript
window._pendingApprovals
```
Should be empty or not contain the approval ID

### Step 4: Verify Execution
- Claude should continue execution
- Ping output should appear in chat

**Console Should Show:**
```
[ApprovalCard] Sending AI-conversational approval as chat message
[ApprovalCard] Sent approval response via WebSocket: yes
```

## Test Scenario 2: Custom Command Modification

### Step 1: Initiate Command
- Type: `run ping google.com -c 100`
- Wait for approval card

### Step 2: Click Custom Instructions
- Click "Custom Instructions" button
- Custom input field should appear

### Step 3: Enter Modified Command
- Type in custom input: `ping google.com -c 5`
- Click "Execute Custom"

**Expected Result:**
- Approval card disappears
- System message: "✅ Approval sent - continuing execution..."
- Custom command sent as WebSocket message

**Console Should Show:**
```javascript
// WebSocket message should be:
{
  type: 'command',
  command: 'ping google.com -c 5',  // Custom command
  metadata: {
    isApprovalResponse: true,
    approvalId: '<approval-id>',
    originalCommand: 'ping google.com -c 100'
  }
}
```

## Test Scenario 3: Reject Command

### Step 1: Initiate Command
- Type: `run rm -rf /` (or any dangerous command)
- Wait for approval card

### Step 2: Click Reject
- Click "Reject" button

**Expected Result:**
- Approval card disappears
- "no" message sent to Claude
- Claude should acknowledge rejection

**Console Should Show:**
```javascript
// WebSocket message:
{
  type: 'command',
  command: 'no',
  metadata: {
    isApprovalResponse: true,
    approvalId: '<approval-id>'
  }
}
```

## Test Scenario 4: Multiple Pending Approvals

### Step 1: Send Multiple Commands
- Type: `run ping google.com`
- Before approving, type: `run ls -la`
- Before approving, type: `run date`

**Expected Result:**
- Multiple approval cards appear
- Each has unique approval ID

**Verify Pending Approvals:**
```javascript
Object.keys(window._pendingApprovals).length
```
Should show 3

### Step 2: Approve Each Individually
- Click "Approve" on first card
- Verify only that card disappears
- Verify other cards remain

**Expected:** Each approval is independent

## Test Scenario 5: WebSocket Disconnected

### Step 1: Disconnect WebSocket
- In DevTools Console, type:
```javascript
window.ws.close()
```

### Step 2: Try to Approve
- Type: `run ping google.com`
- Wait for approval card
- Click "Approve"

**Expected Result:**
- Approval card disappears
- Error message: "❌ Failed to send approval: WebSocket not connected"

**Console Should Show:**
```
[ApprovalCard] WebSocket not connected for approval response
```

### Step 3: Reconnect and Try Again
- Refresh page to reconnect WebSocket
- Send command again
- Approve should work

## Test Scenario 6: Empty Custom Command

### Step 1: Initiate Command
- Type: `run ping google.com`
- Wait for approval card

### Step 2: Try Empty Custom Command
- Click "Custom Instructions"
- Leave custom input empty
- Click "Execute Custom"

**Expected Result:**
- Error message: "Please enter a command"
- Input field regains focus
- Approval card remains visible

## Test Scenario 7: Approval Expiration

### Step 1: Initiate Command
- Type: `run ping google.com`
- Wait for approval card

### Step 2: Wait for Expiration (5 minutes)
- Wait 5 minutes without approving

**Expected Result:**
- Approval card shows "⏱️ Expired"
- Buttons are hidden
- Custom input is hidden

**Note:** This test takes 5 minutes. For faster testing, you can modify the expiration time in server.js temporarily.

## Test Scenario 8: XSS Prevention

### Step 1: Send Command with HTML
- Type: `run <script>alert('XSS')</script>`
- Wait for approval card

### Step 2: Verify Escaping
- Check that command is displayed as text, not rendered HTML
- No alert should appear

**Expected:** Command is escaped using `escapeHtml()` function

## Automated Verification Commands

Run these in DevTools Console to verify the implementation:

```javascript
// 1. Verify ApprovalCard is loaded
console.log('ApprovalCard loaded:', typeof window.ApprovalCard === 'object');

// 2. Verify required methods exist
const methods = ['handleApprove', 'handleReject', 'handleCustom', 'executeCustom', 'sendApprovalResponse', 'handleExpired'];
const hasAllMethods = methods.every(m => typeof window.ApprovalCard[m] === 'function');
console.log('All required methods present:', hasAllMethods);

// 3. Verify pending approvals storage
console.log('Pending approvals initialized:', typeof window._pendingApprovals === 'object');

// 4. Verify WebSocket connection
console.log('WebSocket connected:', window.ws && window.ws.readyState === WebSocket.OPEN);

// 5. Verify session ID
console.log('Session ID:', window.attachedSessionId || window.chatSessionId);
```

## Server-Side Verification

Check server logs for:

```
[WebSocket] Sending command to session <session-id>: yes...
[WebSocket] ✓ Command sent successfully to session <session-id>
```

Verify that the metadata is preserved:
```
metadata: {
  isApprovalResponse: true,
  approvalId: 'approval-<timestamp>-<random>',
  originalCommand: 'ping google.com'
}
```

## Common Issues and Solutions

### Issue 1: Approval card doesn't appear
**Check:**
- Is approval-card.js loaded? (Check Network tab)
- Is ApprovalCard defined? (`console.log(window.ApprovalCard)`)

### Issue 2: Clicking Approve does nothing
**Check:**
- Is WebSocket connected? (`window.ws.readyState === WebSocket.OPEN`)
- Is there a pending approval? (`window._pendingApprovals`)
- Check console for errors

### Issue 3: "yes" message appears in chat instead of executing
**Check:**
- Is the server receiving the command type correctly?
- Check server logs for "Sending command to session"

### Issue 4: Approval appears but Claude doesn't continue
**Check:**
- Did Claude receive "yes" message?
- Check server logs for command delivery
- Verify sessionId matches current session

## Success Criteria

The approval flow fix is working correctly if:

1. ✅ Approval card appears when Claude requests approval
2. ✅ Clicking "Approve" sends "yes" as WebSocket command
3. ✅ WebSocket message includes `isApprovalResponse: true` metadata
4. ✅ Claude receives the "yes" message and continues execution
5. ✅ Custom commands work correctly
6. ✅ Reject button sends "no" message
7. ✅ Pending approvals are cleaned up after response
8. ✅ Multiple approvals can be handled independently
9. ✅ Error handling works for disconnected WebSocket
10. ✅ HTML escaping prevents XSS

## Test Results Checklist

- [ ] Test Scenario 1: Basic Approval Flow - PASS/FAIL
- [ ] Test Scenario 2: Custom Command Modification - PASS/FAIL
- [ ] Test Scenario 3: Reject Command - PASS/FAIL
- [ ] Test Scenario 4: Multiple Pending Approvals - PASS/FAIL
- [ ] Test Scenario 5: WebSocket Disconnected - PASS/FAIL
- [ ] Test Scenario 6: Empty Custom Command - PASS/FAIL
- [ ] Test Scenario 7: Approval Expiration - PASS/FAIL
- [ ] Test Scenario 8: XSS Prevention - PASS/FAIL

**Overall Result:** _____ PASS/FAIL

**Tester Notes:**
_______________________________________________________________
_______________________________________________________________
_______________________________________________________________