Skip to main content
After analyzing your code, ScryCLI can help you fix detected errors through AI-generated code updates. This guide shows you how to leverage AI suggestions for efficient error resolution.

​
How AI-Powered Fixes Work

ScryCLI’s fix workflow:
  1. Detects errors through code analysis
  2. Generates fixes using AI understanding of your codebase
  3. Applies changes through structured file operations
  4. Validates the fixes in context
The AI engine uses your project’s file tree and existing code patterns to generate contextually appropriate fixes.

​
Basic Fix Workflow

1

Identify the Error

First, ask the AI about your code to find issues:
> Check src/hooks/useToolExecutor.ts for error handling issues
The AI will provide a conversational response identifying potential issues, such as missing error handling around JSON.parse.
2

Request AI Fix

Ask ScryCLI to fix the identified issue:
> Add error handling to the JSON.parse in useToolExecutor.ts
Or be more specific:
> Update useToolExecutor.ts to wrap JSON.parse in try-catch and handle parse errors gracefully
3

Review the Fix

ScryCLI will generate a write action:
{
  "action": "write_file",
  "file": "src/hooks/useToolExecutor.ts",
  "content": "// Full updated file content with fix applied"
}
The tool executor will automatically apply the changes to your file.
4

Verify the Fix

After the fix is applied, verify it works:
> Read src/hooks/useToolExecutor.ts and confirm the error handling is correct

​
File Update Actions

ScryCLI supports multiple file modification actions:

​
Write/Modify Existing File

The write_file action modifies an existing file: 
> Fix the authentication bug in src/lib/auth.ts
Response: 
{
  "action": "write_file",
  "file": "src/lib/auth.ts",
  "content": "import { getConfig } from '../config/configManage.js';\n// Updated code with fix..."
}
The write_file action overwrites the entire file content. The AI includes the full file content in the response, not just the changed lines. While the AI’s system prompt mentions update_file, it is processed the same as write_file by the tool executor.

​
Common Fix Patterns

​
Adding Error Handling

# Add try-catch blocks
> Wrap the file operations in src/tools/ with proper error handling

# Add null checks
> Add null checking to the config access in src/lib/isModelSelected.ts

# Handle async errors
> Add error handling to the async llmCall in src/model/openRouter.ts

​
Type Safety Fixes

# Add TypeScript types
> Add proper type definitions to the useChat hook parameters

# Fix type errors
> Fix the TypeScript error in the llmCall return type

# Add type guards
> Add type guards to the JSON parsing in useToolExecutor

​
Security Fixes

# Improve authentication
> Strengthen the JWT verification in src/lib/auth.ts

# Add input validation
> Add path validation to prevent directory traversal in readFile

# Sanitize user input
> Add input sanitization to the file path resolver

​
Advanced Fix Scenarios

​
Multi-File Fixes

When a fix requires changes across multiple files: 
# Step 1: Update the interface
> Update the llmCallParams interface in src/model/openRouter.ts to include options

# Step 2: Update the implementation
> Update the llmCall function to use the new options parameter

# Step 3: Update callers
> Update useChat.ts to pass options when calling llmCall
For complex multi-file fixes, address one file at a time and verify each change before moving to the next.

​
Refactoring Fixes

# Improve code structure
> Refactor the file tree generation to use a class-based approach

# Extract common logic
> Extract the path resolution logic from all tool functions into a shared utility

# Simplify complex functions
> Simplify the useToolExecutor effect by breaking it into smaller functions

​
Configuration Fixes

# Fix config structure
> Update configManage.ts to validate config schema before saving

# Add config defaults
> Add default values for missing config keys in getConfig

# Improve config types
> Add TypeScript interfaces for the config structure

​
Understanding Fix Responses

ScryCLI returns fixes as JSON instructions that the tool executor automatically applies: 
{
  "action": "write_file",
  "file": "src/hooks/useToolExecutor.ts",
  "content": "import { useEffect, useState } from 'react';\n// Full file content..."
}

​
Response Fields

  • action: The operation type (write_file, create_file, read_file, delete_file)
  • file: Relative path to the file
  • content: Complete new file content (for write/create actions)

​
Automatic Execution

The useToolExecutor hook automatically:
  1. Parses the JSON response
  2. Validates the action
  3. Executes the file operation
  4. Handles errors during execution
// From src/hooks/useToolExecutor.ts
switch (instruction.action) {
  case 'write_file':
    writeFile(instruction.file, instruction.content);
    break;
  case 'create_file':
    createFile(instruction.file, instruction.content);
    break;
  // ...
}

​
Best Practices for AI-Assisted Fixes

​
1. Review Before Accepting

Always review AI-generated fixes: 
# Read the file before fix
> Read src/lib/auth.ts

# Apply the fix
> Fix the JWT verification issue

# Read the file after fix
> Read src/lib/auth.ts

​
2. Be Specific About Requirements

Provide clear fix instructions: Good: 
> Add a try-catch block around JSON.parse in useToolExecutor.ts that logs the error and sets a default instruction object
Bad: 
> Fix the error

​
3. Test After Fixes

Verify fixes work in your environment: 
# After applying a fix
> Verify the fix by reading the updated file

# Then manually test
npm run build
npm run dev

​
4. Understand the Context

Make sure the AI has the right context: 
# Provide context for complex fixes
> Update the authentication flow to support OAuth, keeping compatibility with the existing JWT approach in src/lib/auth.ts

​
Example Fix Workflows

​
Fixing a Type Error

# Step 1: Identify
> Check src/model/openRouter.ts for type errors

# AI Response:
# "Property 'model' not found on config"

# Step 2: Fix
> Add optional chaining to config.model access and handle undefined case

# Step 3: Verify
> Read src/model/openRouter.ts and check the fix

# Step 4: Test
npm run build

​
Fixing a Security Issue

# Step 1: Security scan
> Analyze src/tools/readFile.ts for security vulnerabilities

# AI Response:
# "Path traversal vulnerability - user input not validated"

# Step 2: Request fix
> Add path validation to readFile to prevent directory traversal attacks

# Step 3: Review implementation
> Read src/tools/readFile.ts

# Step 4: Apply similar fix to other tools
> Apply the same path validation to writeFile, createFile, and deleteFile

​
Refactoring for Maintainability

# Step 1: Identify code smell
> Review src/hooks/useToolExecutor.ts for code quality issues

# Step 2: Request refactor
> Extract the instruction parsing logic into a separate utility function

# Step 3: Create new file
> Create src/utils/parseInstruction.ts with the extracted logic

# Step 4: Update original
> Update useToolExecutor.ts to use the new parseInstruction utility

​
Handling Fix Failures

​
Invalid JSON Response

If the AI returns invalid JSON: 
# The tool executor will log:
# "Error executing tool: Unexpected token"

# Try rephrasing your request:
> Update src/lib/auth.ts to add better error handling - respond with valid JSON only

​
File Not Found

# If file doesn't exist:
# "Error: ENOENT: no such file or directory"

# Verify the file path:
> Show me the file tree for the src directory

# Then use correct path:
> Update src/lib/auth.ts (not lib/auth.ts)

​
Merge Conflicts

If you’ve manually edited files:
  1. Commit your changes first
  2. Request the AI fix
  3. Manually merge if needed

​
Next Steps

File Management

Learn about file operations through natural language

Code Analysis

Master code analysis techniques