Created and maintained by Varun Kulkarni · ⚡ Quickstart ↓ · Skills Catalog ↓ · Frameworks ↓
Purpose-built, framework-driven skills for designing UX for AI products, agents, and experiences. Each skill encodes a proprietary framework for the unique UX challenges that only exist when humans interact with AI — trust calibration, hallucination recovery, agentic control, prompt interfaces, and more.
# 1. Clone the repo
git clone https://github.com/varunk130/ai-ux-skill-library.git
# 2. Install all 12 skills globally for Claude Code
mkdir -p ~/.claude/skills
cp -r ai-ux-skill-library/skills/* ~/.claude/skills/
# 3. Restart Claude Code, then invoke any framework:
# /ai-conversation-architect — DIALOGUE framework for chat UX
# /ai-trust-transparency — GLASS framework for explainability
# /ai-agent-ux — AUTONOMY framework for agentic UX💡 Full setup including GitHub Copilot integration is in Installation below.
Traditional UX skills don't cover AI. When your product can hallucinate, act autonomously, and produce different outputs from the same input — you need a new UX design vocabulary. This library provides it.
General UX skills (journey mapping, accessibility, design systems) are well-served by existing resources. This library focuses exclusively on the delta — the 12 UX challenges that are unique to AI products and don't exist in traditional software or digital products.
The skills are organized into 3 design phases for AI products:
flowchart TD
F["🎯 FOUNDATION<br/>How users start with AI<br/><br/>ai-onboarding-calibration · ai-prompt-ux · ai-journey-mapper"]
I["🤖 INTERACTION<br/>How users work with AI<br/><br/>ai-conversation-architect · ai-agent-ux<br/>ai-feedback-loops · ai-multimodal-output"]
T["🛡 TRUST & SAFETY<br/>How users trust AI<br/><br/>ai-trust-transparency · ai-error-resilience<br/>ai-safety-guardrails · ai-personalization-ethics"]
F --> I --> T
classDef foundation fill:#1a73e8,color:#fff,stroke:#1558b0,stroke-width:2px,rx:8,ry:8
classDef interact fill:#7C83FD,color:#fff,stroke:#5F65CC,stroke-width:2px,rx:8,ry:8
classDef trust fill:#ea8600,color:#fff,stroke:#c57200,stroke-width:2px,rx:8,ry:8
class F foundation
class I interact
class T trust
12 skills total — 10 core skills (numbered 1–10) plus 2 bonus skills (output/multimodal and accessibility) that extend coverage to rendering and inclusive design.
| # | Skill | Framework | Phase | What It Solves |
|---|---|---|---|---|
| 1 | AI Conversation Architect | DIALOGUE |
Interaction | Conversational AI interfaces — turn-taking, persona voice, multi-turn context, error recovery |
| 2 | AI Trust & Transparency | GLASS |
Trust & Safety | Explainability UX — confidence indicators, citation design, source attribution |
| 3 | AI Error Resilience | RECOVER |
Trust & Safety | Hallucinations, uncertainty, graceful degradation, safe fallbacks |
| 4 | AI Agent UX | AUTONOMY |
Interaction | Agentic AI — autonomy controls, consent, action previews, undo/rollback, audit trails |
| 5 | AI Onboarding & Calibration | CALIBRATE |
Foundation | Progressive disclosure, mental model calibration, expectation setting |
| 6 | AI Feedback Loops | SIGNAL |
Interaction | RLHF UX — thumbs up/down, preference ranking, human-in-the-loop |
| 7 | AI Prompt UX | CRAFT |
Foundation | Prompt interface design — input affordances, templates, suggestions |
| 8 | AI Personalization & Ethics | ADAPT |
Trust & Safety | Adaptive interfaces, privacy balance, filter bubble prevention |
| 9 | AI Safety Guardrails | SHIELD |
Trust & Safety | Content moderation UX, bias detection, harm prevention, refusal design |
| 10 | AI Journey Mapper | PATHWAY |
Foundation | AI-specific journey mapping — trust arcs, capability discovery, autonomy transitions |
| Bonus | AI Output & Multimodal Design | RENDER |
Interaction | Response formatting, output hierarchy, cross-modal presentation |
| Bonus | AI Accessibility Audit | CLEAR |
Trust & Safety | WCAG 2.2 AA audit tailored to AI surfaces — keyboard, screen reader, captions, motion |
| Framework | Mnemonic | Core Concept |
|---|---|---|
| DIALOGUE | _D_iscover, _I_dentify, _A_dapt, _L_ayer, _O_ffer, _G_uard, _U_nderstand, _E_xit | Design conversations, not command lines |
| GLASS | _G_round, _L_ayer, _A_dvertise, _S_how, _S_upport | Make AI reasoning visible |
| RECOVER | _R_ecognize, _E_xpress, _C_ontain, _O_ffer, _V_erify, _E_volve, _R_estore | Treat errors as design material |
| AUTONOMY | _A_ction, _U_ser, _T_iered, _O_bservable, _N_arrated, _O_utcome, _M_emory, _Y_ield | Users supervise, AI executes |
| CALIBRATE | _C_ommunicate, _A_nchor, _L_ayer, _I_nvite, _B_uild, _R_ecalibrate, _A_dapt, _T_rack, _E_volve | Onboarding is calibration, not tutorial |
| SIGNAL | _S_urface, _I_ncentivize, _G_raduate, _N_arrate, _A_ggregate, _L_oop | Feedback is a transaction — close the loop |
| CRAFT | _C_onstrain, _R_eveal, _A_ssist, _F_ormat, _T_each | A blank text box is not a prompt UX |
| ADAPT | _A_gency, _D_ata, _A_lternatives, _P_atterns, _T_ested | Personalization is a power dynamic |
| SHIELD | _S_cope, _H_uman, _I_nform, _E_scalation, _L_og, _D_egrade | Safety and usability are not opposites |
| PATHWAY | _P_erception, _A_utonomy, _T_rust, _H_elp, _W_ow, _A_nxiety, _Y_ield | Map what users BELIEVE, not just what they DO |
| RENDER | _R_ight, _E_asy, _N_avigable, _D_irectly, _E_ditable, _R_eproducible | AI generates output. Humans consume meaning |
| CLEAR | _C_ontrast, _L_abels, _E_quivalents, _A_ssist, _R_esponsive | WCAG 2.2 AA audit tailored to AI surfaces |
ai-onboarding-calibration → ai-prompt-ux → ai-conversation-architect →
ai-trust-transparency → ai-error-resilience → ai-feedback-loops
ai-journey-mapper → ai-agent-ux → ai-safety-guardrails →
ai-trust-transparency → ai-feedback-loops
ai-journey-mapper → ai-trust-transparency → ai-error-resilience →
ai-safety-guardrails → ai-personalization-ethics
ai-onboarding-calibration → ai-journey-mapper → ai-feedback-loops →
ai-personalization-ethics
Each skill is a standalone SKILL.md file that can be installed into your Claude Code or GitHub Copilot environment.
# Clone this repo
git clone https://github.com/varunk130/ai-ux-skill-library.git
# Copy all skills to your Claude Code skills directory
cp -r ai-ux-skill-library/skills/* ~/.claude/skills/
# Or install a single skill
cp -r ai-ux-skill-library/skills/ai-agent-ux ~/.claude/skills/# Clone this repo
git clone https://github.com/varunk130/ai-ux-skill-library.git
# Copy all skills to your GitHub Copilot instructions directory
cp -r ai-ux-skill-library/skills/* .github/skills/
# Or install a single skill
cp -r ai-ux-skill-library/skills/ai-agent-ux .github/skills/- AI-only problems — Every skill targets a UX challenge that does NOT exist in traditional software (trust arcs, hallucination recovery, autonomy dials)
- Proprietary frameworks — Each skill has a named, mnemonic framework (DIALOGUE, GLASS, RECOVER, etc.) with original scoring rubrics and decision matrices
- Anti-pattern catalogs — Every skill includes specific anti-patterns with explanations of why they fail, not just best practices
- Cross-skill integration — Skills reference each other, creating a composable system where outputs from one skill feed into another
- Opinionated defaults — Specific numbers, thresholds, and recommendations rather than "it depends" advice
ai-ux-skill-library/
├── README.md
├── LICENSE
└── skills/
├── ai-conversation-architect/SKILL.md # DIALOGUE Framework
├── ai-trust-transparency/SKILL.md # GLASS Framework
├── ai-error-resilience/SKILL.md # RECOVER Framework
├── ai-agent-ux/SKILL.md # AUTONOMY Framework
├── ai-onboarding-calibration/SKILL.md # CALIBRATE Framework
├── ai-feedback-loops/SKILL.md # SIGNAL Framework
├── ai-prompt-ux/SKILL.md # CRAFT Framework
├── ai-personalization-ethics/SKILL.md # ADAPT Framework
├── ai-safety-guardrails/SKILL.md # SHIELD Framework
├── ai-accessibility-audit/SKILL.md # CLEAR Framework (Bonus, WCAG 2.2 AA)
├── ai-journey-mapper/SKILL.md # PATHWAY Framework
└── ai-multimodal-output/SKILL.md # RENDER Framework (Bonus)
We welcome contributions! To add or improve a skill:
- Fork this repository
- Create a feature branch (
git checkout -b improve-skill-name) - Update the
SKILL.mdin the relevant skill directory - Submit a Pull Request with a description of your changes
Part of a portfolio of AI agent and skill libraries for product, GTM, and decision-making teams.
Discovery & research
- ai-customer-discovery-skills — Turn raw customer signal into validated product opportunities (5 of 12 shipped)
- jtbd-extractor — Extract Jobs-to-be-Done statements from research, with opportunity scoring
Strategy & decisions
- claude-code-skills — 29 production-grade skills for finance, product, strategy, and game theory
- AI-Builder-Decision-Analyst — 11 skills that catch bad bets before you ship across DECIDE / BUILD / COMMUNICATE / LEARN
Go-to-market
- ai-gtm-skill-library — 31 opinionated GTM skills across the full discover → renew lifecycle
- ai-marketing-claude-skills — 12 marketing-ops skills with scoring algorithms and statistical frameworks
- ai-marketing-os — Run a marketing org as an orchestrated AI team: 9 agents, 12 expert skills, end-to-end missions that ship real artifacts
- ai-partner-ecosystem-analysis — Deep research on any ISV, partner, or competitor with a 1-slide PPTX output
Multi-agent demos
- ai-pm-agents-suite — 6-agent pipeline plus 3 standalone PM agents (decision engine, financial analyst, stakeholder translator) that turn customer feedback into strategy, PRDs, and comms
- ai-legal-agents-skills-os — Agentic operating system for legal work: one master agent, nine specialist skills, MCP + MCP Apps
- ai-growth-os — Compound: a multi-agent growth experiment engine that runs the find → design → ship → learn loop end-to-end, offline
- ai-customer-acquisition — Beacon: five agents allocate a budget across channels, build the creative, and reallocate on week-1 results
- ai-revops — Atlas: nine agents across GTM, Partnerships, and RevOps take a company into a new market with one Vertical Launch Plan
Evaluation & operations
- AI-Eval-Skills — 7 skills to plan, generate, run, interpret, and triage AI agent evaluations
- ai-workflow-playbooks — 21 playbooks + 10 skills + 4 guardians + 5 runbooks across the 7-stage delivery pipeline
This project is licensed under the MIT License — see LICENSE for the full text.
Built by Varun Kulkarni
Powered by Claude Code & GitHub Copilot
Released under the MIT License.