Production rollout

Intercept middleware can block, redact, mask, log, or modify prompts before they reach an AI provider. For production apps, it is usually best to roll out gradually.

Recommended rollout path

Start by observing. Then tune. Then enforce.

Local environment

In local development, use logging so you can understand what the middleware detects.

use PromptPHP\Intercept\InjectionGuard\PromptInjectionGuard;
use PromptPHP\Intercept\PIIRedactor\PIIRedactor;

public function middleware(): array
{
    return [
        new PromptInjectionGuard(
            action: 'log',
        ),

        new PIIRedactor(
            action: 'log',
            blockEntities: [],
        ),
    ];
}

This lets prompts continue while writing detection logs.

Staging environment

In staging, continue logging but use traffic and examples that look closer to production.

public function middleware(): array
{
    return [
        new PromptInjectionGuard(
            action: 'log',
        ),

        new PIIRedactor(
            action: 'redact',
            blockEntities: [],
        ),
    ];
}

This lets you confirm that PII redaction works without blocking user flows too early.

Production environment

In production, use stricter defaults.

public function middleware(): array
{
    return [
        new PromptInjectionGuard(
            action: 'block',
        ),

        new PIIRedactor(
            action: 'redact',
            blockEntities: [
                'credit_card',
                'api_key',
                'bearer_token',
            ],
        ),
    ];
}

This setup:

blocks common prompt injection attempts
redacts common structured PII
blocks high-risk secrets
avoids exposing raw sensitive values to the provider

Suggested environment matrix

Environment	Injection Guard	PII Redactor	Notes
Local	`log`	`log`	Good for development and debugging.
Staging	`log`	`redact`	Good for testing real behaviour.
Production	`block`	`redact`	Good default for public agents.
Internal tools	`warn` or `log`	`mask` or `redact`	Depends on trust level and data sensitivity.
High-risk public agents	`block`	`redact` with blocked secrets	Recommended for exposed workflows.

Getting Started

Middleware collection

Contributor resources

Guides & Reference

Updates

Recommended rollout path

Local environment

Staging environment

Production environment

Suggested environment matrix

​Recommended rollout path

​Local environment

​Staging environment

​Production environment

​Suggested environment matrix

Recommended rollout path

Local environment

Staging environment

Production environment

Suggested environment matrix