Skip to content

LinzLos/agentic-trust-ux

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Agentic Trust · Agent Monitor — making invisible trust legible

Live demo · built on the Tiny Wire design system

A concept exploration for the emerging world of agentic commerce: sovereign agents that hold wallets, move money, and run inside confidential hardware you can't see. The guarantees are real but invisible — so this prototype designs the missing UX vocabulary for invisible trust.

Independent concept piece — one worked example in agentic-trust UX.

The problem

The hard question this explores: "How does a user visualize what their agent is doing in the background… without being overwhelmed by technical logs?" — and how do you build the interfaces where humans provision, monitor, and trust autonomous AI agents with their data and their money? That's the brief this prototype answers.

Three primitives

Primitive What it solves
Attestation Shield Translates a TEE attestation quote (Intel TDX / NVIDIA Confidential Computing) into a glanceable "verified enclave" verdict — with the raw cryptographic proof one click away. Glanceable for owners, auditable for engineers.
Trust Dial A radial gauge where trust is set per action, not per app. Drag the threshold and watch real scenarios ($4 inference vs. a $42k rebalance) fall into the ambient or pre-flight zone. Friction scales with stakes.
Intent timeline + pre-flight Shows what the agent is doing (declarative intent, competing solvers, gas abstraction) instead of logs — raw trace one toggle away. High-stakes intents gate behind a pre-flight with manual levers (spend cap, pause-after).

Where this fits the stack

  • Agent runtime → agent monitor + pre-flight approval flow
  • Confidential cloud / GPU market → the Attestation Shield (TEEs, attestable compute)
  • Agentic UI & Intents → legible timeline, declarative pre-flight, solver resolution
  • Design system → built entirely on semantic tokens, so a single brand-token change re-flows everywhere

Design stance

  • Legibility over logs. Intent first; the machine trace is never the default.
  • Intervenability over automation-only. Pause, cap, and override mid-flight — always.
  • Glanceable but auditable. The Shield reads in a second; the proof stays reachable for the skeptic.
  • Web2 ease, Web3 sovereignty. Keys stay yours; friction scales with the stakes.

Click "Designer's cut" in the top bar to reveal the Decision / Why / Trade-off behind each.

Run it

No build step. It's plain HTML/CSS/JS on the Tiny Wire design system.

python3 -m http.server 4317
# open http://localhost:4317

Built on Tiny Wire

Every surface uses Tiny Wire semantic tokens (--brand, --surface, --text-*) and components — no hardcoded colors. The design system is the stable contract between design and engineering; this prototype is a consumer of it. See CREDITS.

About

Concept work on agentic trust UX — making cryptographic attestation and autonomous agent activity legible and auditable. An interactive agent-monitor concept with a per-action Trust Dial and a pre-flight decision surface. Built on the Tiny Wire design system.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors