Adversary
CatalogSDKDocsCLI
CLISign In
⌘K
OverviewGetting startedCLICI / GitHub ActionsPlatform componentsSDKRuntimeAdversariesPermissionsDistributionRegistry authenticationExecution trust modelOfficial signatures

Documentation

⌘K
OverviewGetting startedCLICI / GitHub ActionsPlatform componentsSDKRuntimeAdversariesPermissionsDistributionRegistry authenticationExecution trust modelOfficial signatures

Platform components

Authoring, execution, distribution, and trust are separate layers. Teams include a private hosted namespace.

The public catalog is free (login for usage tracking). Every team also gets a private hosted namespace for internal adversaries, with role-based push and pull access. External OCI registries remain supported.

CLI

Open-source binary to search, pull, run, pack, push, and sign adversaries. Primary interface for local development and CI.

SDK

Open-source TypeScript SDK for rules, observations, findings, and the runtime review envelope.

Runtime

How adversaries execute: review scope, permissions, model access, reporting, and host vs future sandboxes.

Public catalog

Free official packages on registry.adversarylabs.ai under domain/name ids (go/security, ci/github-actions, …). Browse on the site; pull with the CLI.

Official signatures

Ed25519 signatures over content digests. Release CLI verifies them for host-execution trust.

Documentation

Ask AI

Grounded in the local docs index. Answers can include links to the matching pages.

Ask how to install the CLI, authenticate, run adversaries, or understand platform concepts.

Enter to send · Shift+Enter for a new line

Adversary

Free open catalog of executable AI adversaries. Sign in to pull for usage tracking. Run locally. Keep private artifacts on any OCI registry.

$ brew install adversarylabs/tap/adversary

$ adversary run go/security --path .

Product

  • Catalog
  • Getting started
  • CLI
  • CI / GitHub Actions
  • Docs

Platform

  • SDK
  • Runtime
  • Distribution
  • Teams

Resources

  • Documentation
  • Official catalog
  • Trust model
© 2026 Ambiguity Labs LLC. All rights reserved.adversary v1