error-handling-confirmation

name	error-handling-confirmation
description	This skill normalizes errors and produces human-friendly confirmations across API and UI layers. It prevents silent failures, ensures graceful degradation, and maintains consistent UX quality. Use when implementing error responses, success confirmations, "not found" handling, or improving agent response quality.
allowed-tools	Read, Grep, Glob, Edit, Write

Error Handling & Confirmation

What This Skill Does

• Normalizes error responses into consistent, human-friendly formats
• Produces clear success confirmations that build user trust
• Handles "not found" and edge cases gracefully
• Prevents silent failures through explicit error surfacing
• Ensures agent/API responses maintain UX quality

When to Use

• Implementing API error responses
• Creating user-facing error messages
• Adding success/confirmation feedback
• Handling "task not found" or similar scenarios
• Building fallback responses for AI agents
• Improving response quality and polish

When NOT to Use

• Logging/monitoring (use observability patterns)
• Exception handling architecture (use language-specific patterns)
• Input validation logic (handle separately, use these patterns for messages)

---

Core Principles

1. Never Blame the User

❌ "You entered an invalid email"
✅ "Please enter a valid email address"

❌ "You don't have permission"
✅ "This action requires additional permissions"

2. Be Specific and Actionable

❌ "An error occurred"
✅ "Unable to save task. Please check your connection and try again."

❌ "Invalid input"
✅ "Title must be less than 200 characters (currently 215)"

3. Use Plain Language

❌ "Error 0x80004005: Unspecified error"
✅ "Something went wrong. Please try again."

❌ "ECONNREFUSED 127.0.0.1:5432"
✅ "Unable to connect to the database. Please try again later."

4. Preserve User Trust

❌ Silent failure (no feedback)
✅ "We couldn't complete that action. Your data is safe."

❌ Technical stack trace
✅ "Something unexpected happened. Our team has been notified."

---

Error Response Patterns

API Error Structure

# Standardized error response
{
    "error": {
        "code": "TASK_NOT_FOUND",      # Machine-readable
        "message": "Task not found",    # Human-readable
        "details": {}                   # Optional context
    }
}

HTTP Status Code Mapping

Scenario	Status	Code	Message
Resource not found	404	NOT_FOUND	"{Resource} not found"
Access denied	403	FORBIDDEN	"You don't have access to this {resource}"
Auth required	401	UNAUTHORIZED	"Please sign in to continue"
Validation failed	400	VALIDATION_ERROR	Specific field message
Server error	500	INTERNAL_ERROR	"Something went wrong. Please try again."
Service unavailable	503	SERVICE_UNAVAILABLE	"Service temporarily unavailable"

Not Found Handling (Critical Pattern)

Distinguish between "doesn't exist" and "not yours":

def get_task(task_id: int, user_id: str):
    task = db.get_task(task_id)

    if task is None:
        # Task doesn't exist at all
        raise NotFoundError("Task not found")

    if task.user_id != user_id:
        # Task exists but belongs to someone else
        raise ForbiddenError("You don't have access to this task")

    return task

Why this matters: Returning 404 for "not yours" leaks information about resource existence.

---

Confirmation Patterns

Success Messages

Action	Confirmation
Create	"Task created successfully"
Update	"Changes saved"
Delete	"Task deleted"
Complete	"Nice work! Task completed"
Bulk action	"3 tasks updated"

Confirmation Components

// Toast notification pattern
addToast("Task created successfully", "success");
addToast("Unable to save changes", "error");
addToast("Are you sure?", "warning");

Visual Feedback Hierarchy

1. Immediate - Button state change (loading → success)
2. Transient - Toast notification (auto-dismisses)
3. Persistent - Inline message (stays until resolved)
4. Blocking - Modal dialog (requires action)

---

Graceful Degradation

Fallback Strategy

Primary action fails
    ↓
Try recovery (retry, alternative)
    ↓
If still fails → Show friendly error
    ↓
Preserve user's work/state
    ↓
Offer clear next steps

Never Lose User Data

// Bad: Silent failure
async function saveTask(task) {
    await api.save(task);  // If fails, data lost
}

// Good: Preserve on failure
async function saveTask(task) {
    try {
        await api.save(task);
        showSuccess("Task saved");
    } catch (error) {
        // Keep form data intact
        showError("Couldn't save. Your changes are preserved.");
        // Optionally save to localStorage as backup
    }
}

Agent/Chatbot Fallbacks

User input not understood
    ↓
"I didn't quite catch that. Could you rephrase?"
    ↓
Still unclear after 2 attempts
    ↓
"I'm having trouble understanding. Here's what I can help with: [options]"
    ↓
Offer human escalation if available

---

Message Templates

Error Messages

const ERROR_MESSAGES = {
    // Network
    NETWORK_ERROR: "Unable to connect. Please check your internet connection.",
    TIMEOUT: "Request timed out. Please try again.",

    // Auth
    SESSION_EXPIRED: "Your session has expired. Please sign in again.",
    UNAUTHORIZED: "Please sign in to continue.",
    FORBIDDEN: "You don't have permission to do this.",

    // Resources
    NOT_FOUND: "{resource} not found.",
    ALREADY_EXISTS: "A {resource} with this name already exists.",

    // Validation
    REQUIRED_FIELD: "{field} is required.",
    TOO_LONG: "{field} must be less than {max} characters.",
    INVALID_FORMAT: "Please enter a valid {field}.",

    // Server
    SERVER_ERROR: "Something went wrong. Please try again.",
    SERVICE_UNAVAILABLE: "Service temporarily unavailable. Please try again later.",

    // Generic fallback
    UNKNOWN: "An unexpected error occurred. Please try again."
};

Success Messages

const SUCCESS_MESSAGES = {
    // CRUD
    CREATED: "{resource} created successfully.",
    UPDATED: "Changes saved.",
    DELETED: "{resource} deleted.",

    // Actions
    COMPLETED: "Nice work! {resource} completed.",
    SAVED: "Saved.",
    SENT: "Sent successfully.",

    // Bulk
    BULK_UPDATE: "{count} items updated.",
    BULK_DELETE: "{count} items deleted."
};

---

Implementation Checklist

API Layer

• Consistent error response structure
• Appropriate HTTP status codes
• Human-readable messages (no technical jargon)
• Distinguish 403 vs 404 properly
• No sensitive data in error responses

UI Layer

• Toast system for transient feedback
• Inline validation messages
• Loading states during operations
• Error boundaries for crashes
• Empty states for no-data scenarios

User Experience

• No silent failures
• Clear next steps in error messages
• User data preserved on failure
• Consistent tone (friendly, not blaming)
• Appropriate visual feedback (icons, colors)

---

Anti-Patterns

Don't Do This

Anti-Pattern	Problem	Solution
Silent failure	User confused	Always show feedback
Technical jargon	User can't understand	Plain language
Generic "Error"	Not actionable	Be specific
Blame language	Damages trust	Use neutral tone
Stack traces	Security risk, confusing	Log server-side only
-1 or null returns	Easy to miss	Throw explicit errors
Alert boxes	Disruptive UX	Use toasts/inline

Security Considerations

# Bad: Leaks information
"User [e-mail] not found"  # Confirms email doesn't exist

# Good: Generic auth messages
"Invalid email or password"  # Same message for both cases

---

Before Implementation

Gather context to ensure successful implementation:

Source	Gather
Codebase	Existing error handling, toast system, message patterns
Conversation	User's specific requirements, tone preferences
Skill References	Domain patterns from references/
User Guidelines	Team's preferred messaging style

---

Integration with Other Skills

• api-client-design: Error response structure integration
• frontend-architecture: Toast and notification patterns
• fastapi-architecture: HTTPException handling
• auth-aware-ui: Auth error flow handling

---

References

See references/ for:

• error-catalog.md - Complete error code catalog
• message-templates.md - Copy-paste message templates
• ux-patterns.md - Visual feedback patterns