Registry

Available Agents

Each agent is a complete operational package — inspect the dossier, understand the authority boundaries, and take one home.

Privacy Champion

Standalone agent Experimental

"Audit and harden telemetry surfaces across your OS, browsers, and network."

Overview

Fingerprints your system, enumerates known telemetry and data-exfiltration channels, probes current state, classifies risk, presents findings for approval, applies closures, verifies, and produces a complete audit trail with rollback instructions.

Designed for first-time computer setup, quarterly privacy hygiene, pre-travel preparation, and anyone who wants to understand what their machine is sharing — and with whom.

40+ Surfaces catalogued
3 Platforms
2 Taskrails
v0.1.0 Version

Authority Model

Capability Behaviour
Read Automatic — fingerprinting, probing, classifying are safe and run freely
Propose Automatic — findings presented with risk classification and tiered recommendations
Mutate OS settings Explicit approval per surface — every toggle requires individual sign-off
Mutate browser settings Explicit approval per surface
Mutate network config Explicit approval per surface
Destructive actions Prohibited by default — escalates with warning and requires additional confirmation
Escalation Halt and report — never infers approval from context

Platform Compatibility

OS Versions Browsers
Windows 10, 11 Chrome, Firefox, Edge, Brave
macOS 13 (Ventura), 14 (Sonoma), 15 (Sequoia) Chrome, Firefox, Safari, Edge, Brave
Linux Ubuntu 22.04+, Fedora 39+, Debian 12+ Chrome, Firefox, Edge, Brave

Package Structure

The agent lives at agents/privacy-champion/ in the TAO repository, with two taskrails (8-step privacy audit + 7-step induction), five SOP scripts, a surface catalog of 40+ known telemetry channels, and a full local adaptation workflow.

Good Fit

  • First-time setup of a new personal computer
  • Quarterly privacy hygiene audit
  • Suspicion that a system update re-enabled telemetry
  • Pre-travel privacy preparation for crossing borders
  • Journalists, researchers, and others with elevated threat models
  • Anyone who wants to understand what their computer is sharing

Poor Fit

  • Corporate-managed or MDM-enrolled devices (group policy may override changes)
  • Servers or headless systems (different threat model)
  • Mobile devices (iOS/Android telemetry differs significantly)
  • Users unwilling to review and approve changes before they are applied
  • Air-gapped or fully offline systems

Coming soon

More agents in development. If you have a use case or want to contribute an agent package, open an issue on the repository .