auxfirst — agentic UX as a named discipline.
Patterns, heuristics, and trust systems for AI products. We don't design interfaces. We design relationships between users and agents.
Project home: auxfirst.com · aux-audit · Schemas · Onboarding · Docs · Cite this work
AUX (Agentic User Experience) is an open standard. The 10 AUX Heuristics, the 4-stage Trust Architecture, and the Trust Gap Taxonomy are published as editable YAML schemas — forkable, citable, not vendor-locked. The schemas are MIT — embed them, vendor them, no attribution condition. The prose around them is CC BY 4.0.
If you ship AI products, you are no longer designing screens. You are designing an ongoing relationship between a user and an agent that has memory, initiative, and judgment. Classical UX was built for tools. AUX is built for relationships.
The standard is executable. aux-audit scores an agent product against the 10 heuristics and the Trust Architecture, and tells you which trust stage you have actually earned.
# from a clone of this repo
npm --prefix packages/aux-audit ci && npm --prefix packages/aux-audit run build
node packages/aux-audit/dist/cli.js run ./agent-spec.yaml
# after npm publish
npx aux-audit run ./agent-spec.yamlScore 41/100 · Grade F · Trust stage: functional
| Stage | Earned | Blocked by |
| 1. Functional Trust | ✅ | — |
| 2. Contextual Trust | ❌ | aux.H08 |
In CI, as a GitHub Action — SARIF into the Security tab, a scorecard on the PR:
- uses: auxfirst/trustkit@v0.2
with:
spec: ./agent-spec.yaml
fail-on: highIt grades the spec, not the running product: it can prove a mechanism was never declared, never that a declared one works. A spec with no evidence of testing is capped at present on every heuristic, so the score cannot be gamed with better prose.
Under agent-spec v1 it computes the gap between what an agent's credentials permit and what its mandate governs — the attack surface — and treats a row enforced by "the system prompt" as unenforced. See packages/aux-audit.
Pick the role that's most you. Read five minutes. Walk away with one artifact you can forward.
| Who you are | Read this | Walk away with |
|---|---|---|
| Business operator / GTM | for-business-operators | 10 questions for any AI vendor |
| CTO / VP Engineering | for-ctos | CTO one-pager |
| Engineer shipping the agent | for-engineering | aux-audit CI workflow |
| Solo / indie builder | two-patterns | Two patterns, copy-paste |
Or read onboarding/README.md — the router with all four side by side, plus appendices and principles.
.
├── onboarding/ ← audience-routed onboarding (R1–R4 + forwardables A1–A4)
│ ├── README.md ← the router
│ ├── for-business-operators.md ← R1 (3 pages, plain English)
│ ├── for-ctos.md ← R2 (4 pages, peer-to-peer technical)
│ ├── for-engineering.md ← R3 (4 pages, code-heavy)
│ ├── two-patterns.md ← R4 (3 pages, copy-paste patterns)
│ ├── appendix-a-vocabulary.md ← AUX vocabulary
│ ├── appendix-b-index.md ← project file index
│ ├── principles.md ← the five principles
│ ├── traps.md ← common ways this goes wrong
│ ├── auxfirst-onboarding-v0.2.docx ← R1–R4 as one Word bundle
│ └── forwardables/ ← standalone take-aways A1–A4
│
├── schemas/ ← canonical YAML — the standard
│ ├── aux-heuristics.yaml ← the 10 heuristics
│ ├── trust-architecture.yaml ← the 4 trust stages
│ ├── trust-gap-taxonomy.yaml ← named failure modes
│ ├── aux-evolution-curve.yaml ← the 4 capability stages (teardowns only)
│ ├── trust-contract.yaml ← example Trust Contract
│ ├── agent-spec.schema.yaml ← v1.0, real JSON Schema — per-action mandate
│ ├── agent-spec.v0.yaml ← v0.1.0, frozen; still scored, not comparable
│ ├── MIGRATION.md ← v0 → v1, and why
│ ├── migrate-v0-to-v1.py ← converter; leaves TODOs rather than guessing
│ ├── brain-spec.schema.yaml ← Agency Brain input format (JSON Schema)
│ └── memory-policy.schema.yaml ← memory governance format
│
├── repos/ ← first-wave repo READMEs (split out as v0.x ships)
│ ├── aux-frameworks/ ← Definition
│ ├── aux-audit/ ← Executable (spec; built in packages/aux-audit)
│ ├── trust-gap-classifier/ ← Executable
│ ├── agent-memory-policy/ ← Executable
│ ├── agent-ux-teardowns/ ← Reference
│ └── agentic-ux-patterns/ ← Reference (intent-handshake, memory-in-motion)
│
├── packages/aux-audit/ ← the CLI (MIT) · npx aux-audit
├── action.yml ← GitHub Action wrapping the CLI
│
├── visuals/storyboard.html ← single-page visual overview
├── one-pager/ ← business one-pager (Word, editable)
├── profile/README.md ← the github.com/auxfirst org profile
│
├── ARCHITECTURE.md ← big-picture system view
├── REPO_STRUCTURE.md ← naming + layout conventions
├── CONTRIBUTING.md ← contribution bar (the bar is high)
├── CODE_OF_CONDUCT.md ← Contributor Covenant 2.1
├── docs/ ← static-site copy (canonical path TBD)
├── CHANGELOG.md ← what shipped
├── CITATION.cff ← how to cite AUX definitions
├── LICENSE ← CC BY 4.0 (prose; GitHub-detectable legal code)
├── LICENSING.md ← license map: machine-readable = MIT, prose = CC BY
└── .github/ ← Issue templates, PR template, CI
- Random experiments, SaaS clones, tutorials, or think-pieces dressed as code.
- Opinions without benchmarks.
- Frameworks without schemas.
One technical post with runnable code + benchmark numbers will out-earn ten soft think-pieces.
This is the proof layer for the category. Every claim ships with a schema, a CLI, or a reproducible scenario.
- onboarding/README.md — the router. Pick a slice, skip the rest.
- packages/aux-audit — the CLI. Run it against your own spec.
- docs/index.md — the static-site copy of the same material.
- ARCHITECTURE.md — how the pieces fit.
- visuals/storyboard.html — download and open in a browser; the whole story on one page.
- repos/ — the six first-wave repos.
- schemas/ — the YAML that makes this real.
- one-pager/auxfirst-github-onepager.docx — for business conversations.
- GitHub = receipts for AUX, not a projects page.
- Every repo is exactly one of three kinds: Definition, Executable, Reference.
- Every claim has a runnable example.
- Name before you tool.
- The internal execution layer and the public standard layer share one taxonomy.
The bar is high. Patterns need diagrams, heuristics need examples, benchmarks need a runnable script. See CONTRIBUTING.md before opening a PR. Open an Issue first for anything that touches an ID or introduces vocabulary — five Issue templates are set up: pattern-request, gap-proposal, debate, schema, bug.
The definitions here are meant to be quoted, forked, and built on. If you reference the AUX Heuristics, the Trust Architecture, or the Trust Gap Taxonomy in a post, paper, or product doc, cite the repository — CITATION.cff carries the machine-readable metadata, and GitHub renders it behind the Cite this repository button as APA or BibTeX.
auxfirst (2026). TrustKit — AUX (Agentic User Experience) heuristics, trust architecture, and trust gap taxonomy (Version 0.2). https://github.com/auxfirst/trustkit
One rule: machine-readable definitions are MIT, prose is CC BY 4.0.
- Schemas (
schemas/*.yaml), pattern examples (example.py), and tooling (packages/aux-audit,action.yml): MIT. Embed them; no attribution condition. - Prose (onboarding, vocabulary, pattern write-ups and diagrams): CC BY 4.0.
- Teardown content: CC BY-NC 4.0 (attribution, non-commercial).
GitHub's badge reads CC BY 4.0 because it detects the root LICENSE, which covers the prose. The schemas are not CC — see LICENSING.md and schemas/LICENSE.
v0.2 · 2026-04-19 — First public release. Onboarding layer live: audience-routed docs (R1–R4) + four forwardable artifacts + two canonical patterns + six first-wave repo READMEs + canonical schemas. aux-audit v0.1 is implemented, tested, and wired into CI as a GitHub Action. Seven patterns close all 12 named trust gaps, enforced in CI. Next: npm publish so npx aux-audit resolves without a clone; list the Action on the Marketplace; name the Evolution Curve stages for teardowns (#5 — the axis is human-assessed, not something aux-audit computes).
TrustKit — auxfirst. Agentic UX as a named discipline.
auxfirst.com · github.com/auxfirst · questions and proposals go in Issues.