laminae-shadow — Adversarial Red-Teaming Engine

The Shadow is an automated security auditor that red-teams AI output. It runs as an async post-processing pipeline — never blocking the user's conversation — and produces structured vulnerability reports.

Pipeline Stages

1. Static analysis — regex pattern scanning (always runs)
2. LLM adversarial review — local Ollama model with attacker-mindset prompt
3. Sandbox execution — ephemeral container testing (optional)

Each stage implements the [Analyzer] trait and can be extended or replaced.

Quick Start

use laminae_shadow::{ShadowEngine, ShadowEvent, create_report_store};

#[tokio::main]
async fn main() {
    let store = create_report_store();
    let engine = ShadowEngine::new(store.clone());

    let mut rx = engine.analyze_async(
        "session-1".into(),
        "Here's some code:\n```python\neval(user_input)\n```".into(),
    );

    while let Some(event) = rx.recv().await {
        match event {
            ShadowEvent::Finding { finding, .. } => {
                println!("[{}] {}: {}", finding.severity, finding.category, finding.title);
            }
            ShadowEvent::Done { report, .. } => {
                println!("Analysis complete: {}", report.summary);
            }
            _ => {}
        }
    }
}