UX Mob Process

The UX Mob Process plugin brings AI-native UX elaboration into Claude Code as a structured, human-gated workflow. It guides product teams through phased design sessions — from domain analysis to handoff artifacts — enforcing strict approval gates, context isolation, and governance guardrails at every step. No artifact is saved and no phase advances without explicit human approval.

It supports two project categories: Greenfield (starting from scratch) and Brownfield (evolving an existing product), with a mandatory AI Readiness assessment before any brownfield work begins.

Standalone installation — no Xianix required

This plugin can be installed directly into your local Claude Code environment. You do not need to configure it through Xianix at all. See Local Installation below to get started. The Xianix Agent Studio Integration section at the bottom of this page is only relevant if you are also running the Xianix platform.

---

What You Get Back

Each completed session produces a set of structured artifacts saved to your project workspace under [workspace-root]/projects/[project-folder]/:

Artifact	Greenfield	Brownfield
domain.md	Phase 1	AI Readiness Phase 3
personas.md / user-groups-and-roles.md	Phase 2	AI Readiness Phase 4
executable-product-intent.md	Phase 3	—
client-input.md, intent-gap.md, client-verification.md	Phases 4–6	—
feature-map.md	Phase 9	AI Readiness Phase 5
user-journeys.md / current-user-journeys.md	Phase 10	AI Readiness Phase 7
units.md + unit-validation-cases.md	Phases 11–12	Post-AI-Readiness work types
design.md	Phase 13	—
outputs/ui-prompts/	Phase 14	—
greenfield-handoff.md / brownfield work type artifacts	Phase 16	Post-readiness
decisions/ (assumptions register, decision log, do-not-invent list, drift register)	All phases	All phases
outputs/ai-readiness/ (11 readiness artifacts)	—	AI Readiness

Every artifact is previewed in chat before saving. Saving an artifact does not approve the phase — they are separate explicit gates.

---

How It Works

flowchart TD
    A["/ux-start"] --> B{Project type?}
    B -->|Greenfield| C[Load greenfield process]
    B -->|Brownfield| D[Run AI Readiness — mandatory]
    C --> E[Phase-by-phase elaboration\nwith approval gates]
    D --> F{AI Readiness\napproved?}
    F -->|Yes| G[Select brownfield work type\nAdd Feature / Revamp / Improve]
    G --> H[Execute work type phases\nwith approval gates]
    E --> I[Artifacts saved to workspace\nafter explicit human approval]
    H --> I
    I --> J["/ux-status — check progress\n/ux-resume — continue later\n/ux-approve-phase — advance phases"]

Greenfield (16 phases):

1. Domain Analysis — domain.md: terminology, industry context, regulatory patterns.
2. Persona Analysis — personas.md: user groups, goals, pain points.
3. Executable Product Intent — the “why” and outcomes behind the product.
4. Client Input Capture — structured capture of client-provided requirements.
5. Intent Gap Analysis — gaps between client input and product intent.
6. Client Verification & Scope Agreement — explicit scope sign-off with the client.
7. Intent Cleanup / Scope Removal — removes out-of-scope items from active artifacts.
8. MVP Scope Suggestion (optional) — suggests a minimal viable scope cut.
9. Feature Map — prioritized feature list with in/out-of-scope status.
10. User Journeys — end-to-end flows per persona across features.
11. AI-DLC Unit Definition — product units derived from approved feature map and journeys.
12. Unit Validation Cases — traceability cases from each unit to a product source.
13. Design Direction — visual language, component guidance, and UI personality.
14. UI Prompt Generation — one AI design prompt per user journey.
15. Design Review & Corrective Prompts (optional) — reviews generated designs against heuristics and validation cases.
16. Handoff — consolidated handoff document for implementation.

Brownfield (AI Readiness first — 11 phases, then work type):

The AI Readiness phase captures the existing product context before any changes begin: artifact audit, product brief, domain, user groups & roles, feature map, navigation map, current user journeys, design system context, product rules/constraints/do-not-invent list, AI context pack, and readiness approval. Only after readiness is approved can a brownfield work type (Add New Feature / Complete Revamp / Improve Existing Feature) begin.

---

Commands

Command	What it does
/ux-start [greenfield | brownfield]	Initialize a new UX Mob session. Prompts for project type, name, and workspace path.
/ux-start-fresh	Start with strict context isolation — guarantees no prior run outputs or archived artifacts leak in.
/ux-resume [project-folder]	Resume a paused session from the last approved phase in project-state.json.
/ux-status [project-folder]	Show the current project status: active phase, completed phases, saved artifacts, and open questions.
/ux-run-ai-readiness	Run (or re-run) the Brownfield AI Readiness assessment standalone. Required before any brownfield work type begins.
/ux-approve-phase	Record explicit phase approval after running the Phase Quality Checklist and advance to the next phase.
/ux-validate-guardrails	Run a full governance compliance check and output a Guardrail Validation Report.
/ux-validate-context-drift	Check whether any unapproved context from prior runs has leaked into the active session.
/ux-run-matrix	Run only the UX Matrix module to generate a prioritized feature-persona matrix from approved artifacts.

---

Governance & Guardrails

The plugin enforces a strict governance framework throughout every session. These rules cannot be bypassed silently — violations cause the agent to stop and ask the human how to proceed.

Strict No-Silent-File-Write Rule — the agent must never write an artifact to disk without first showing the full proposed content in chat and receiving explicit save approval. This applies to every file in the project workspace.

Artifact Save ≠ Phase Approval — saving a file and approving a phase are two separate explicit confirmations. A human can save all drafts in a phase and still revise before marking the phase complete.

Context Isolation Rule — each new session treats the project as isolated. Prior mob runs, archived artifacts, old conversation memory, and unapproved files from outside the active project folder may not be used without explicit human authorization.

Template Enforcement Rule — before generating any artifact, the agent reads the relevant template and preserves its heading structure exactly. Missing content is marked Needs human input, assumptions as Assumption, and inferred content as Inferred.

Mandatory Brownfield AI Readiness — brownfield projects must complete and receive human approval on AI Readiness before a work type (Add New Feature, Complete Revamp, Improve Existing Feature) may begin.

Long-Artifact Preview Rule — for large artifacts, the agent must offer section-by-section preview rather than dumping the full content in one block. The human reviews and approves each section before the artifact is assembled and saved.

Do-Not-Invent Rule — the agent may not invent domain knowledge, business rules, user context, or product decisions. If information is missing, it must ask the human or mark it explicitly rather than filling in the gap.

---

Project Workspace Layout

When a session is initialized, the plugin creates this structure outside the plugin directory:

[workspace-root]/projects/[project-folder]/
├── project-state.json                     # Phase state, approvals, artifact registry
├── outputs/
│   ├── *.md                               # Greenfield artifacts (domain, personas, etc.)
│   ├── ai-readiness/                      # Brownfield AI Readiness artifacts
│   ├── ui-prompts/                        # One UI prompt file per user journey
│   └── corrective-prompts/                # Design review corrective prompts
├── phase-artifacts/                       # Phase-specific artifacts (e.g., ux-matrix output from /ux-run-matrix)
├── decisions/
│   ├── do-not-invent-list.md
│   ├── assumptions-register.md
│   ├── decision-log.md
│   ├── ai-drift-register.md
│   └── guardrail-validation-report.md
└── archives/                              # Prior run artifacts (brownfield revivals)

Note

The plugin never writes inside its own directory. All project artifacts go to the workspace root you specify at session start.

---

Sample Session

# Start a new Greenfield project
/ux-start greenfield

# Check where you are at any point
/ux-status

# After reviewing and approving a phase
/ux-approve-phase

# Pause and come back later
/ux-resume my-project-folder

# Run a compliance check on all saved artifacts
/ux-validate-guardrails

For a Brownfield project, AI Readiness runs automatically first:

/ux-start brownfield
# → AI Readiness runs for 11 phases before work type is selected

# Or run AI Readiness standalone on an existing project
/ux-run-ai-readiness

---

Local Installation

Install the plugin directly into your Claude Code environment — no Xianix configuration needed.

One-time setup (choose one):

# Option A: copy to your Claude Code plugins directory
cp -r /path/to/ux-mob-process-plugin ~/.claude/plugins/ux-mob-process-plugin

# Option B: register it in your .claude.json
# Add "ux-mob-process-plugin" to the "plugins" array pointing at the folder path

Reload Claude Code, then verify the plugin loaded:

/ux-status
# Should respond: [ux-mob-process] loaded

Start a session:

# Greenfield session
/ux-start greenfield

# Brownfield session (AI Readiness runs first)
/ux-start brownfield

You will be prompted for a project name, project folder name, and workspace root path. The plugin initializes the workspace and begins the first phase immediately.

---

Plugin Architecture

The plugin is composed of agents, skills, and a session hook. You don’t invoke these directly — the commands wire them together automatically.

Agents:

Agent	Role
ux-mob-facilitator	Primary orchestrator. Routes to Greenfield or Brownfield process, enforces approval gates and governance rules throughout.
greenfield-process-agent	Executes all 16 Greenfield phases with per-phase routing and special rules.
brownfield-ai-readiness-agent	Runs the 11-phase AI Readiness assessment and unlocks brownfield work type selection on completion.
design-review-agent	Executes the optional Phase 15 design review using Nielsen’s 10 usability heuristics and generates corrective prompts.
context-drift-validator-agent	Read-only auditor. Validates context isolation and produces a Context Drift Report.
artifact-auditor-agent	Read-only auditor. Checks draft artifacts against template structure, source declarations, and labeling requirements before save approval.

Skills:

Skill	Role
ux-mob-process-runner	Core execution loop: project initialization, phase routing, approval gates, artifact saves, and resume-from-state.
brownfield-ai-readiness	Guides the 11-phase AI Readiness flow with artifact existence checks and reconstruction rules.
ux-guardrail-validator	Full compliance validation across phase sequencing, source-of-truth chains, scope rules, and labeling requirements.

Hook: On session start, the plugin prints [ux-mob-process] loaded — run '/ux-start' to begin a Greenfield or Brownfield UX Mob session. — a quick confirmation that the plugin is active.

---

Xianix Agent Studio Integration

This section is only relevant if you are running the Xianix platform. If you installed the plugin locally into Claude Code, you can ignore this entirely.

Add the execution block below to your rules.json to run UX Mob sessions directly inside an Agent Studio chat conversation.

Execution-block shape

Field	Purpose
name	Human-readable identifier for this rule
match-any	Trigger filter — x-xianix-chat-trigger? activates the rule from an Agent Studio chat
use-plugins	The plugin to load, referenced by name and marketplace URL
model	Claude model to use for the session
max-budget-usd	Maximum spend cap for the session
max-turns	Maximum conversation turns before the session closes
execute-prompt	The opening prompt sent to the agent when the session starts

Sample rules.json

{
  "name": "chat-ux-mob-process",
  "match-any": [
    {
      "name": "chat-only-stub",
      "rule": "x-xianix-chat-trigger?"
    }
  ],
  "use-plugins": [
    {
      "plugin-name": "ux-mob-process@xianix-plugins-official",
      "marketplace": "xianix-team/plugins-official"
    }
  ],
  "model": "claude-sonnet-4-5",
  "max-budget-usd": 2,
  "max-turns": 150,
  "execute-prompt": "Run /ux-mob-process:ux-start to begin the UX Mob process. Ask the user whether this is a Greenfield or Brownfield project if not already stated."
}

Note

The UX Mob Process is designed as an interactive, human-gated workflow. A human must remain present in the session to review drafts and approve each phase. The agent will not proceed autonomously past any approval gate.