Crate ironclaw_safety 

Safety layer for prompt injection defense.

This crate provides protection against prompt injection attacks by:

• Detecting suspicious patterns in external data
• Sanitizing tool outputs before they reach the LLM
• Validating inputs before processing
• Enforcing safety policies
• Detecting secret leakage in outputs

Structs

InjectionWarning

Warning about a potential injection attempt.

LeakDetector

Detector for secret leaks in output data.

LeakMatch

A detected potential secret leak.

LeakPattern

A pattern for detecting secret leaks.

LeakScanResult

Result of scanning content for leaks.

Policy

Safety policy containing rules.

PolicyRule

A policy rule that defines what content is blocked or flagged.

SafetyConfig

Safety configuration.

SafetyLayer

Unified safety layer combining sanitizer, validator, and policy.

SanitizedOutput

Result of sanitizing external content.

Sanitizer

Sanitizer for external data.

ValidationResult

Result of validating input.

Validator

Input validator.

Enums

LeakAction

Action to take when a leak is detected.

LeakDetectionError

Error from leak detection.

LeakSeverity

Severity of a detected leak.

PolicyAction

Action to take when a policy is violated.

Severity

Severity level for safety issues.

Functions

params_contain_manual_credentials

Check whether HTTP request parameters contain manually-provided credentials.

wrap_external_content

Wrap external, untrusted content with a security notice for the LLM.