Approve what automation can touch
File, network, desktop automation, inter-applet messaging, and other sensitive actions are bounded by the permissions approved for that applet or workflow.
YAAA.app is designed for AI-assisted, human-approved automation that runs local-first, admits only approved capabilities, and denies protected packages when the user, entitlement, or package integrity check does not match.
This page is informational and is not a formal security whitepaper. Do not submit secrets, credentials, proprietary scripts, or vulnerability details through public website channels until a verified disclosure path is live.
Applets and automation run under explicit, least-privilege permissions. An applet asks for capabilities, the runtime admits only the capabilities requested and granted, and denied capabilities fail closed before dispatch or protected access.
File, network, desktop automation, inter-applet messaging, and other sensitive actions are bounded by the permissions approved for that applet or workflow.
Capabilities can be scoped, pruned, or rejected. A workflow can run only inside the permission envelope it was granted, instead of receiving broad ambient access.
Unsupported, incomplete, unauthorized, or tampered applet packages are refused. Sensitive operations are designed around explicit approval instead of silent escalation.
Plugins and applet packages are protected before they run. Packages are encrypted with AES-256-GCM and cryptographically signed with Ed25519, then verified, authorized, and decrypted only when the required checks pass.
A package with a bad signature, invalid identity, unsupported package data, or changed encrypted contents is denied instead of loaded.
An applet or plugin author can require a signed entitlement so only the authenticated users the developer designates can decrypt and run the protected package.
Protected-package keys are not shipped as plaintext files beside the package. Missing credentials, missing entitlement, the wrong user, or a tampered package all fail closed.
Current verified encryption covers applet packages, plugin packages, and production configuration. Production project-file encryption is an active-development control and should be evaluated as part of release readiness before sensitive deployments.
The AI companion is being built around app-owned controls, OAuth provider authentication, local-first execution, and human-approved actions. Provider connections are explicit, and requests fail closed when required credentials are missing.
Claude and OpenAI provider connections are handled through authenticated provider configuration, so AI access is explicit instead of assumed.
High-impact actions are framed around previews, confirmations, app-owned commands, and audit evidence. The AI suggests and operates through controlled surfaces; the user approves sensitive work.
Applets and project work are designed to live and run on the user's machine or chosen infrastructure by default. External AI, license, update, telemetry, or storage connections should be explicit and documented before use.
YAAA.app cannot protect projects from a fully compromised workstation. Keep operating systems, drivers, dependency managers, and endpoint controls patched before relying on sensitive automation on a local machine.
Review privacy, licensing, pricing, downloads, and support paths before deploying YAAA.app in a sensitive environment.
Privacy policy → · License terms → · Pricing → · Download checklist → · Contact support →