AI-Adversary app
AI-Adversary brings a unified dashboard, API, and CLI to authorized security engagements, target management, approvals, and assessment results. Its agents investigate web applications, APIs, networks, source code, dependencies, and infrastructure configuration with tools including ZAP, Semgrep, Trivy, and Checkov.
Every proposed action passes deterministic checks for scope, current permissions, impact, approvals, and budgets before execution. Durable job state and execution records support restart recovery, avoid duplicate work where outcomes are known, and keep cancellation or unresolved cleanup visible.
Authenticated API testing compares users, object ownership, response content, and secure controls to distinguish authorization flaws from misleading responses. Isolated browser and HTTP proxy runtimes support navigation, forms, traffic inspection, and bounded replay with screenshots and request-response evidence.
Findings connect observations, verification decisions, and versioned artifacts to reproducible tests with positive and negative controls. Remediation workflows generate patches against an exact source revision, rerun verification and regression controls, and present the diff and evidence for review.
Bounded parallel agents investigate complementary hypotheses under shared scope and budget limits, with coverage tracking for tested, blocked, inconclusive, and untested areas. Diff-scoped CI workflows, SARIF exports, model-usage accounting, and repeatable benchmarks help teams act on findings and assess detection quality, reliability, speed, and cost.
