Rust Guards Guide

The Guard Trait

All guards implement the Guard trait:

pub trait Guard: Send + Sync {
    fn name(&self) -> &str;
    fn check(&self, content: &str) -> GuardCheckResult;
    fn action(&self) -> GuardAction;
    fn severity_threshold(&self) -> Severity { Severity::Low }
}

Using the Derive Macro

The #[derive(Guard)] macro auto-generates the trait implementation.

Field Detection

The macro inspects your struct fields:

Field	Type	Generated Method
name	String	fn name() -> &str { &self.name }
action	GuardAction	fn action() -> GuardAction { self.action }
severity_threshold	Severity	fn severity_threshold() -> Severity { self.severity_threshold }

Missing fields get defaults: - No name → uses snake_case of struct name - No action → defaults to GuardAction::Block - No severity_threshold → defaults to Severity::Low

Attributes

Use #[guard(name = "...")] to set a fixed name:

#[derive(Guard)]
#[guard(name = "my_fixed_name")]
struct MyGuard {
    action: GuardAction,
}

The guard_check Method

You must implement guard_check(&self, content: &str) -> GuardCheckResult. The macro's generated check() delegates to this method:

impl MyGuard {
    fn guard_check(&self, content: &str) -> GuardCheckResult {
        // Your logic here
    }
}

GuardCheckResult

Use the convenience constructors:

// Content is safe
GuardCheckResult::pass("guard_name")

// Content is flagged
GuardCheckResult::fail("guard_name", GuardAction::Block, matches, "reason")

// With sanitized content
GuardCheckResult::fail(...).with_sanitized("cleaned content".to_string())

GuardAction Options

Action	Description
Allow	Content passes through
Block	Content is blocked (default)
Sanitize	Content is cleaned/redacted
Log	Content is logged but allowed
Alert	Alert is raised but content allowed
Suggest	Suggestion provided, no blocking

Severity Levels

Level	Score	Use Case
Info	0	Informational only
Low	1	Minor issues
Medium	2	Standard threats (default)
High	3	Significant threats
Critical	4	Severe threats requiring immediate action

Utilities

The utils module provides text processing helpers:

use oxide_guard_sdk::utils;

let tokens = utils::tokenize("Hello, World!");
// ["hello", "world"]

let normalized = utils::normalize("  Hello   World  ");
// "hello world"

let matched = utils::pattern_match("Hello World", "hel*rld");
// true

let score = utils::similarity("hello world", "hello earth");
// 0.333...