convez.ai
Skill
varunk130

ai-ux-skill-library

varunk130/ai-ux-skill-library

Ver no GitHub
BomPluginScore · de 0 a 100

The 12-Skill AI UX Design Engine for Claude Code & GitHub Copilot — purpose-built skills for designing UX for AI products, AI agents, and AI-powered experiences

Instalação

$npx convez-ai varunk130/ai-ux-skill-library

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
3
Forks
0
Acompanhando
0
Saúde
85%
Último commit
há 1 semana
Versão
—
Issues abertas
0
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

AI UX Skill Library

The 12-Skill AI UX Design Engine for Claude Code & GitHub Copilot

Skills Frameworks License: MIT Built with Claude Code

Created and maintained by Varun Kulkarni · ⚡ Quickstart ↓ · Skills Catalog ↓ · Frameworks ↓

Purpose-built, framework-driven skills for designing UX for AI products, AI agents, and AI-powered 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.


⚡ Quickstart

# 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.


Why This Exists

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.


Ecosystem Architecture

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

Skills Catalog

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.

#SkillFrameworkPhaseWhat It Solves
1AI Conversation ArchitectDIALOGUEInteractionConversational AI interfaces - turn-taking, persona voice, multi-turn context, error recovery
2AI Trust & TransparencyGLASSTrust & SafetyExplainability UX - confidence indicators, citation design, source attribution
3AI Error ResilienceRECOVERTrust & SafetyHallucinations, uncertainty, graceful degradation, safe fallbacks
4AI Agent UXAUTONOMYInteractionAgentic AI - autonomy controls, consent, action previews, undo/rollback, audit trails
5AI Onboarding & CalibrationCALIBRATEFoundationProgressive disclosure, mental model calibration, expectation setting
6AI Feedback LoopsSIGNALInteractionRLHF UX - thumbs up/down, preference ranking, human-in-the-loop
7AI Prompt UXCRAFTFoundationPrompt interface design - input affordances, templates, suggestions
8AI Personalization & EthicsADAPTTrust & SafetyAdaptive interfaces, privacy balance, filter bubble prevention
9AI Safety GuardrailsSHIELDTrust & SafetyContent moderation UX, bias detection, harm prevention, refusal design
10AI Journey MapperPATHWAYFoundationAI-specific journey mapping - trust arcs, capability discovery, autonomy transitions
BonusAI Output & Multimodal DesignRENDERInteractionResponse formatting, output hierarchy, cross-modal presentation
BonusAI Accessibility AuditCLEARTrust & SafetyWCAG 2.2 AA audit tailored to AI surfaces - keyboard, screen reader, captions, motion

Framework Quick Reference

FrameworkMnemonicCore Concept
DIALOGUE_D_iscover, _I_dentify, _A_dapt, _L_ayer, _O_ffer, _G_uard, _U_nderstand, _E_xitDesign conversations, not command lines
GLASS_G_round, _L_ayer, _A_dvertise, _S_how, _S_upportMake AI reasoning visible
RECOVER_R_ecognize, _E_xpress, _C_ontain, _O_ffer, _V_erify, _E_volve, _R_estoreTreat errors as design material
AUTONOMY_A_ction, _U_ser, _T_iered, _O_bservable, _N_arrated, _O_utcome, _M_emory, _Y_ieldUsers supervise, AI executes
CALIBRATE_C_ommunicate, _A_nchor, _L_ayer, _I_nvite, _B_uild, _R_ecalibrate, _A_dapt, _T_rack, _E_volveOnboarding is calibration, not tutorial
SIGNAL_S_urface, _I_ncentivize, _G_raduate, _N_arrate, _A_ggregate, _L_oopFeedback is a transaction - close the loop
CRAFT_C_onstrain, _R_eveal, _A_ssist, _F_ormat, _T_eachA blank text box is not a prompt UX
ADAPT_A_gency, _D_ata, _A_lternatives, _P_atterns, _T_estedPersonalization is a power dynamic
SHIELD_S_cope, _H_uman, _I_nform, _E_scalation, _L_og, _D_egradeSafety and usability are not opposites
PATHWAY_P_erception, _A_utonomy, _T_rust, _H_elp, _W_ow, _A_nxiety, _Y_ieldMap what users BELIEVE, not just what they DO
RENDER_R_ight, _E_asy, _N_avigable, _D_irectly, _E_ditable, _R_eproducibleAI generates output. Humans consume meaning
CLEAR_C_ontrast, _L_abels, _E_quivalents, _A_ssist, _R_esponsiveWCAG 2.2 AA audit tailored to AI surfaces

Quickstart Workflows

Designing a New AI Chat Product

ai-onboarding-calibration → ai-prompt-ux → ai-conversation-architect →
ai-trust-transparency → ai-error-resilience → ai-feedback-loops

Designing an AI Agent Experience

ai-journey-mapper → ai-agent-ux → ai-safety-guardrails →
ai-trust-transparency → ai-feedback-loops

Auditing an Existing AI Product

ai-journey-mapper → ai-trust-transparency → ai-error-resilience →
ai-safety-guardrails → ai-personalization-ethics

Improving AI Adoption & Retention

ai-onboarding-calibration → ai-journey-mapper → ai-feedback-loops →
ai-personalization-ethics

Installation

Each skill is a standalone SKILL.md file that can be installed into your Claude Code or GitHub Copilot environment.

Claude Code

# 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/

GitHub Copilot

# 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/

What Makes These Skills Unique

  1. AI-only problems - Every skill targets a UX challenge that does NOT exist in traditional software (trust arcs, hallucination recovery, autonomy dials)
  2. Proprietary frameworks - Each skill has a named, mnemonic framework (DIALOGUE, GLASS, RECOVER, etc.) with original scoring rubrics and decision matrices
  3. Anti-pattern catalogs - Every skill includes specific anti-patterns with explanations of why they fail, not just best practices
  4. Cross-skill integration - Skills reference each other, creating a composable system where outputs from one skill feed into another
  5. Opinionated defaults - Specific numbers, thresholds, and recommendations rather than "it depends" advice

Directory Structure

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)

Contributing

We welcome contributions! To add or improve a skill:

  1. Fork this repository
  2. Create a feature branch (git checkout -b improve-skill-name)
  3. Update the SKILL.md in the relevant skill directory
  4. Submit a Pull Request with a description of your changes

Related Work

Part of a portfolio of AI agent and skill libraries for product, GTM, and decision-making teams.

Discovery & research

Strategy & decisions

Go-to-market

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

Evaluation & operations

  • AI-Eval-Skills - 6 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

License

This project is licensed under the MIT License — see LICENSE for the full text.


Built by Varun Kulkarni

Powered by Claude Code & GitHub Copilot

Ver o original no GitHub

Bom para

Relacionados