agentsclimarketplace

User experience

Skill geekatron/jerry/skills/user-experience

Parent orchestrator for AI-augmented UX methodology targeting tiny teams (1-5 people). Routes to 10 sub-skills by product lifecycle stage through criteria-gated waves. Invoke when team needs structured UX evaluation, user research, design systems, UX metrics, behavior diagnosis, feature prioritization, design sprints, or AI interaction design. Each sub-skill implements a proven UX framework with synthesis hypothesis confidence gates and MCP design tool integration. Triggers: UX, user experience, usability, heuristic evaluation, JTBD, lean UX, HEART metrics, atomic design, inclusive design, behavior design, Kano model, design sprint, AI-first design, UX audit, accessibility, design system, user research.From its SKILL.md

Install
npx -y skills add geekatron/jerry --skill user-experience

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.

SKILL.md

45.9 KB, as published. Nobody here has run it

<!-- NOTE: activation-keywords (19 entries) intentionally differ from mandatory-skill-usage.md trigger map. activation-keywords are SKILL.md-scoped terms for Claude Code agent discovery (H-26). The trigger map uses broader/narrower terms with negative keywords for routing disambiguation (H-36b). This asymmetry is by design per agent-routing-standards.md Enhanced Trigger Map specification: trigger map keywords are routing-optimized; activation-keywords are discovery-optimized. -->

User-Experience Skill

Version: 1.0.0 Framework: Jerry User-Experience Constitutional Compliance: Jerry Constitution v1.0 SSOT References: GitHub Issue #138 (Architecture Spec), PROJ-022 PLAN.md, ADR-agent-design-001 (Agent Definition Format) Project: PROJ-022 User Experience Skill | GitHub Issue #138

Document Sections

SectionPurpose
Document AudienceTriple-Lens audience guide
PurposeSkill overview and key capabilities
When to Use This SkillActivation triggers and scope boundaries
Available Agents11-agent roster with roles, models, and output locations
P-003 ComplianceOrchestrator-worker hierarchy diagram
Invoking an AgentThree invocation methods with examples
Wave Architecture5-wave criteria-gated deployment model
Lifecycle-Stage RoutingOrchestrator triage by product stage
Synthesis Hypothesis Validation3-tier confidence gate protocol
MCP Integration ArchitectureDependency matrix, cost tiers, fallback paths
Cross-Skill IntegrationIntegration with eng-team, red-team, adversary, orchestration, and others
Constitutional ComplianceGoverning principles
Quick ReferenceCommon workflows and agent selection hints
ReferencesFull repo-relative paths to all referenced files

Document Audience (Triple-Lens)

This SKILL.md serves multiple audiences:

LevelAudienceSections to Focus On
L0 (Stakeholder)New users, product managersPurpose, When to Use This Skill, Quick Reference
L1 (Developer)Engineers invoking agentsInvoking an Agent, Available Agents, Lifecycle-Stage Routing
L2 (Architect)Workflow designers, skill maintainersWave Architecture, MCP Integration Architecture, Synthesis Hypothesis Validation

Purpose

The User-Experience skill provides AI-augmented UX methodology for tiny teams (1-5 people) through 11 agents (1 orchestrator + 10 framework specialists). Each agent encapsulates a proven UX framework, produces persistent artifacts that survive context compaction, enforces methodological rigor via synthesis hypothesis confidence gates, and builds a cumulative product UX knowledge base.

The skill targets teams where UX is a part-time responsibility -- calibrated for 20-50% of one person's time in Waves 1-2, with Waves 3-5 as aspirational progression.

Key Capabilities

  • Heuristic Evaluation -- Nielsen's 10 Heuristics with severity-rated findings (Wave 1)
  • Jobs-to-Be-Done Analysis -- User motivation research, switch interviews, job mapping (Wave 1)
  • Lean UX Facilitation -- Hypothesis-driven build-measure-learn cycles (Wave 2)
  • HEART Metrics -- Google's Goals-Signals-Metrics (GSM) framework for UX measurement (Wave 2)
  • Atomic Design -- Brad Frost's 5-level component taxonomy (Wave 3)
  • Inclusive Design Evaluation -- WCAG 2.2 compliance and Microsoft Inclusive Design (Wave 3)
  • Behavior Diagnosis -- Fogg B=MAP bottleneck analysis (Wave 4)
  • Kano Feature Prioritization -- Must-be / Performance / Attractive classification (Wave 4)
  • Design Sprint Facilitation -- AJ&Smart Design Sprint 2.0 four-day process (Wave 5)
  • AI-First Interaction Design -- AI interface patterns and trust calibration (Wave 5, CONDITIONAL)

Deployment Status (v1.0.0): The skill architecture is fully specified but sub-skills deploy incrementally through the wave model. Currently deployed: ux-orchestrator (Wave 0 Foundation). Wave 1-5 sub-skill agents are not yet deployed — invoking the skill before their wave completes will route to only the currently-available sub-skills. The orchestrator discloses deployment state to users per P-022 and offers the wave bypass mechanism for urgent needs. See Wave Architecture for the full deployment timeline.


When to Use This Skill

Activate when:

  • Evaluating an existing interface against usability heuristics
  • Conducting user research to understand motivations (JTBD, switch interviews)
  • Designing or validating a Lean UX hypothesis and experiment cycle
  • Defining or measuring UX metrics using the HEART framework
  • Building or auditing a component library using atomic design principles
  • Auditing for accessibility compliance (WCAG 2.2, inclusive design)
  • Diagnosing why users fail to take a desired action (Fogg B=MAP)
  • Prioritizing features by user satisfaction impact (Kano analysis)
  • Running a structured design sprint for rapid prototyping
  • Designing AI-powered interfaces or conversational interactions
  • Performing a comprehensive multi-framework UX audit (orchestrator routes to 2+ agents)
  • Urgent UX triage via crisis mode (3-skill emergency sequence)

Do NOT use for:

  • Security-focused code review or threat modeling (use /eng-team)
  • Systems engineering requirements or V&V (use /nasa-se)
  • General research without UX focus (use /problem-solving)
  • Adversarial quality review of deliverables (use /adversary)
  • Offensive security testing (use /red-team)
  • Architecture design without user-facing concerns (use /architecture)
  • Documentation writing (use /diataxis)

Available Agents

AgentRoleTierModeModelWaveOutput Location
ux-orchestratorParent orchestrator: routing, wave gating, cross-framework synthesisT5IntegrativeOpus0projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-orchestrator-{topic-slug}.md
ux-heuristic-evaluatorNielsen heuristic evaluation specialistT4SystematicHaiku*1projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-heuristic-evaluator-{topic-slug}.md
ux-jtbd-analystJobs-to-Be-Done research and analysisT4DivergentSonnet1projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-jtbd-analyst-{topic-slug}.md
ux-lean-ux-facilitatorLean UX hypothesis and experiment facilitationT4SystematicSonnet2projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-lean-ux-facilitator-{topic-slug}.md
ux-heart-analystHEART metrics framework specialistT2SystematicSonnet2projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-heart-analyst-{topic-slug}.md
ux-atomic-architectAtomic design component taxonomy architectT4SystematicSonnet3projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-atomic-architect-{topic-slug}.md
ux-inclusive-evaluatorInclusive design and accessibility auditorT4SystematicSonnet3projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-inclusive-evaluator-{topic-slug}.md
ux-behavior-diagnosticianFogg B=MAP behavior bottleneck diagnosisT2ConvergentSonnet4projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-behavior-diagnostician-{topic-slug}.md
ux-kano-analystKano model feature classification and prioritizationT2ConvergentSonnet4projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-kano-analyst-{topic-slug}.md
ux-sprint-facilitatorAJ&Smart Design Sprint 2.0 facilitationT4SystematicOpus5projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-sprint-facilitator-{topic-slug}.md
ux-ai-design-guideAI-first interaction design specialist (CONDITIONAL)T4DivergentOpus5projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-ai-design-guide-{topic-slug}.md

Tool tier key: T2 = Read-Write (Read, Write, Edit, Glob, Grep, Bash); T4 = External (T2 + WebSearch, WebFetch, Context7 MCP, Memory-Keeper); T5 = Orchestration (T4 + Agent). See agent-development-standards.md [Tool Security Tiers] for full definitions. All tier assignments follow the principle of least privilege (AR-006) — T2 agents (ux-heart-analyst, ux-behavior-diagnostician, ux-kano-analyst) operate on user-provided data only; T4 agents access external UX standards and documentation.

*Haiku for high-volume checklist evaluation; escalates to Sonnet when: (1) heuristic severity is "critical" (>= 3 critical findings), (2) Figma MCP benchmark fails pre-launch threshold, or (3) evaluation spans > 50 screens. Escalation is automatic within the orchestrator's routing logic per AD-M-009 model selection justification.

All agents produce output at three levels per AD-M-004:

  • L0 (Executive Summary): Key findings in 3-5 bullets for stakeholders and cross-framework synthesis.
  • L1 (Technical Detail): Full methodology execution with evidence, design recommendations, and component specifications.
  • L2 (Strategic Implications): Cross-product patterns, organizational UX recommendations, and design evolution trajectory.

P-003 Compliance

The /user-experience skill enforces strict single-level nesting per H-01/P-003. Only ux-orchestrator has Agent tool access. All 10 sub-skill agents are workers that MUST NOT include Agent in their tool list.

MAIN CONTEXT (user request)
    |
    v
ux-orchestrator (T5, Opus, Integrative) -- routes, gates, synthesizes
    |
    +-- ux-heuristic-evaluator    (T4, Systematic, Haiku)     [Wave 1]
    +-- ux-jtbd-analyst           (T4, Divergent, Sonnet)     [Wave 1]
    +-- ux-lean-ux-facilitator    (T4, Systematic, Sonnet)    [Wave 2]
    +-- ux-heart-analyst          (T2, Systematic, Sonnet)    [Wave 2]
    +-- ux-atomic-architect       (T4, Systematic, Sonnet)    [Wave 3]
    +-- ux-inclusive-evaluator    (T4, Systematic, Sonnet)    [Wave 3]
    +-- ux-behavior-diagnostician (T2, Convergent, Sonnet)    [Wave 4]
    +-- ux-kano-analyst           (T2, Convergent, Sonnet)    [Wave 4]
    +-- ux-sprint-facilitator     (T4, Systematic, Opus)      [Wave 5]
    +-- ux-ai-design-guide        (T4, Divergent, Opus)       [Wave 5 COND]

Enforcement: Sub-skill agents declare disallowedTools: [Agent] in .md frontmatter. CI gate validates no sub-skill agent has Agent access (documented in skills/user-experience/rules/ci-checks.md). Each .governance.yaml includes Agent prohibition in capabilities.forbidden_actions with P-003 consequence statement.


Invoking an Agent

Option 1: Natural Language Request

Describe your UX need; the orchestrator routes to the appropriate agent:

"Evaluate this dashboard design against Nielsen's heuristics"
"Map the jobs to be done for our onboarding flow"
"Set up a Lean UX experiment for the new search feature"
"Define HEART metrics for the checkout experience"
"Audit this form for WCAG 2.2 AA compliance"
"Run a design sprint for the new mobile navigation"
"Comprehensive UX audit of the settings page"
"CRISIS: users are abandoning checkout -- urgent UX triage"

Option 2: Explicit Agent Request

Request a specific agent by name:

"Use ux-heuristic-evaluator to audit the navigation patterns"
"Have ux-kano-analyst classify our backlog features"
"I need ux-inclusive-evaluator to review color contrast and screen reader compatibility"

Option 3: Native Agent Invocation (Agent Tool)

The orchestrator invokes sub-skill agents as named subagents via Agent:

Agent(
    description="ux-heuristic-evaluator: Heuristic evaluation of settings page",
    subagent_type="ux-heuristic-evaluator",
    prompt="""
## UX CONTEXT (REQUIRED)
- **Engagement ID:** UX-0001
- **Topic:** Settings Page Heuristic Evaluation
- **Product:** [product name and domain]
- **Target Users:** [user description]

## TASK
Perform a Nielsen heuristic evaluation of the settings page.
Evaluate all 10 heuristics. Rate severity 0-4 for each finding.
Produce ranked findings with remediation recommendations.
""",
)

Claude Code enforces each agent's tools frontmatter -- worker agents only have access to their declared tool tier (T2 or T4).


Wave Architecture

Sub-skills deploy in 5 criteria-gated waves. Waves are deployment phases for incremental skill build-out, not runtime execution order. At runtime, the orchestrator routes to any deployed sub-skill based on user need.

Wave Definitions

WaveNameSub-SkillsEntry CriteriaBypass Condition
0Foundationux-orchestrator + rules + templatesPROJ-022 plan approvedN/A
1Zero-Dependency/ux-heuristic-eval, /ux-jtbdKICKOFF-SIGNOFF.md completed with MCP ownership assignmentsN/A (first wave)
2Data-Ready/ux-lean-ux, /ux-heart-metricsWave 1: at least 1 heuristic eval completed AND 1 JTBD job statement used in a product decision2 sprint cycles elapsed with no Wave 1 completion; documented rationale
3Design System/ux-atomic-design, /ux-inclusive-designWave 2: launched product with analytics OR 1 completed Lean UX hypothesis cycleStorybook already in use (skip Lean UX prerequisite for Atomic Design)
4Advanced Analytics/ux-behavior-design, /ux-kano-modelWave 3: Storybook with 5+ Atom stories AND 1 Persona Spectrum reviewExisting user base with analytics (skip Persona Spectrum prerequisite)
5Process Intensives/ux-design-sprint, /ux-ai-first-design (COND)Wave 4: 30+ users for Kano survey OR 1 B=MAP bottleneck diagnosed; AI-First: Enabler DONE + WSM >= 7.80Design Sprint can proceed without Kano prerequisite if team has existing user research

Design Sprint early-access note: Design Sprint is placed in Wave 5 because it requires facilitation infrastructure (Figma + Miro REQ MCPs) and benefits from prior framework exposure. However, teams at product inception ("don't know what to build") who need Design Sprint immediately can use the wave bypass mechanism (3-field documentation: unmet criterion = "no prior wave completion", impact assessment = "sprint proceeds without prior framework calibration", remediation = "backfill Wave 1-2 sub-skills post-sprint"). The lifecycle-stage router handles this: the "Before design: Need validated prototype" route maps to /ux-design-sprint regardless of wave state — the orchestrator presents the bypass prompt if Wave 5 is not yet deployed. This ensures P-020 compliance: the user decides whether to proceed with bypass, not the system.

Wave Transition Quality Gates

Each wave transition is a quality checkpoint. The orchestrator checks WAVE-{N}-SIGNOFF.md existence before routing to Wave N+1 sub-skills.

TransitionQuality CheckThreshold
Wave 0 to 1KICKOFF-SIGNOFF.md completenessAll fields populated (pass/fail)
Wave 1 to 2Wave 1 deliverables quality scoringS-014 composite >= 0.85 on heuristic eval report
Wave 2 to 3Wave 2 deliverables + usage evidenceS-014 composite >= 0.85 + documented usage artifact
Wave 3 to 4Wave 3 deliverables + Storybook artifactS-014 composite >= 0.85 + Storybook story count verification
Wave 4 to 5Wave 4 deliverables + user data evidenceS-014 composite >= 0.85 + user count or behavioral data artifact

Threshold justification (0.85 vs H-13 0.92): Wave transition gates assess sub-skill deployment readiness — whether a sub-skill produces useful output for end users. This is distinct from H-13's 0.92 threshold, which governs C2+ deliverable quality (agent definitions, rule files, governance artifacts). The 0.85 threshold reflects that wave gates evaluate operational output quality rather than governance artifact quality. Formal threshold derivation is tracked in ADR-PROJ022-002-wave-criteria-gates.md (pending); the threshold may be revised upward as calibration data from Wave 1 deployments becomes available.

Wave bypass: Requires 3-field documentation (unmet criterion, impact assessment, remediation plan with target date). Bypass state produces a warning banner on all sub-skill outputs from the bypassed wave. Cumulative ceiling: Maximum 2 concurrent bypasses per team. If a team has 2 active bypasses, the orchestrator requires remediation of at least one before granting additional bypasses. This prevents accumulation of technical UX debt through unbounded wave skipping.

Wave Signoff Enforcement

  • WAVE-N-SIGNOFF.md is a closure deliverable -- wave completion is not recognized until the signoff file passes schema validation (all required fields non-empty) and is committed to the repository
  • Wave transitions tracked via /worktracker entities
  • Templates provided: skills/user-experience/templates/kickoff-signoff-template.md, skills/user-experience/templates/wave-signoff-template.md

Lifecycle-Stage Routing

The orchestrator routes requests through a multi-stage triage based on product lifecycle stage:

1. ONBOARD: Display HIGH RISK user research warning (first invocation per session)
2. CAPACITY CHECK: Ask team UX time allocation
   -> If < 20% of one person's time: recommend Wave 1 sub-skills only (P-020: user decides)
3. MCP CHECK: Detect MCP availability
   -> If MCP unavailable: route to non-MCP fallback paths
4. STAGE TRIAGE: Route by product stage
   |
   +-- "Before design: Don't know what to build"     -> /ux-jtbd
   +-- "Before design: Need to prioritize features"   -> /ux-kano-model
   +-- "Before design: Need validated prototype"        -> /ux-design-sprint
   +-- "During design: Iterating on existing design"   -> /ux-lean-ux OR /ux-heuristic-eval
   +-- "During design: Building component system"      -> /ux-atomic-design
   +-- "During design: Building AI product"            -> /ux-ai-first-design (if Enabler DONE)
   |                                                     OR /ux-heuristic-eval + PAIR (interim)
   +-- "After launch: Measure UX health"               -> /ux-heart-metrics
   +-- "After launch: Users not completing action"     -> /ux-behavior-design
   +-- "Any stage: Check accessibility"                -> /ux-inclusive-design
   +-- "CRISIS: Urgent UX problems"                    -> Emergency 3-skill sequence:
                                                          Heuristic Eval -> Behavior Design -> HEART

Dispatch logic: The orchestrator implements lifecycle-stage routing as a 4-step sequential triage within its <methodology> section. Each step is deterministic: (1) ONBOARD fires once per session via session state flag, (2) CAPACITY CHECK uses a single structured question per P-020, (3) MCP CHECK probes Context7 availability via a lightweight resolve call, (4) STAGE TRIAGE matches user intent against the route table above. If STAGE TRIAGE produces multiple matches (e.g., "iterate on our accessible checkout"), the orchestrator applies the ordering protocol from agent-routing-standards.md [Multi-Skill Combination]: content before quality, work before presentation. If ambiguity remains after ordering, the orchestrator asks the user per H-31 rather than guessing. The CRISIS path bypasses normal triage and executes a fixed 3-skill sequence; the user confirms entry into CRISIS mode but does not select individual sub-skills (P-020 compliance: user authorizes the emergency sequence, orchestrator selects the fixed route). Full CRISIS mode behavior will be specified in the ux-orchestrator agent <methodology> section (EPIC-001).

Common Intent-to-Route Resolution

User SaysRoutes ToQualification Question
"Improve my UX" / "Make this more usable"Heuristic Eval (existing design) or Design Sprint (no design yet)"Do you have an existing design?"
"Fix a specific UX problem"Behavior Design (behavioral) or Heuristic Eval (design-level)"Is the problem about user behavior or design quality?"
"Decide what to build"JTBD (strategic) or Kano (prioritize known features)"Are you defining the problem or prioritizing features?"
"Measure whether UX is working"HEART MetricsNo qualification needed
"Make this accessible"Inclusive DesignNo qualification needed
"CRISIS: urgent UX problems"Emergency 3-skill sequenceNo qualification needed

Routing logic is documented in skills/user-experience/rules/ux-routing-rules.md.


Synthesis Hypothesis Validation

Multiple sub-skills produce AI-generated abstractions ("synthesis hypotheses") that may reflect training data biases rather than the team's specific users. A 3-tier confidence gate fires at skill invocation time.

Confidence Gate Protocol

ConfidenceGate BehaviorAdvancement Rule
HIGHUser reviews output + acknowledges specific AI judgment calls via Synthesis Judgments SummaryAdvances to design decisions after enumerated acknowledgment
MEDIUMRequires expert review OR validation against 2-3 real user data pointsCannot advance to design decisions without named validation source
LOWOutput permanently labeled reference-only; design recommendation section structurally omittedCannot be overridden by any user action

Sub-Skill Synthesis Output Map

Sub-SkillSynthesis StepTypical Confidence
/ux-jtbdJob statement synthesis from secondary researchMEDIUM
/ux-lean-uxAssumption mapping; hypothesis generationMEDIUM
/ux-design-sprintDay 4 interview thematic analysisHIGH
/ux-design-sprintDay 2 sketch selection rationaleMEDIUM
/ux-inclusive-designPersona Spectrum customizationMEDIUM
/ux-kano-modelDirectional classification (5-8 respondents)MEDIUM
/ux-kano-modelFeature priority conflict interpretationLOW
/ux-behavior-designB=MAP bottleneck diagnosisMEDIUM
/ux-behavior-designDesign intervention recommendationLOW
/ux-heart-metricsGoal-metric mapping interpretationMEDIUM
/ux-heart-metricsMetric threshold recommendationLOW
/ux-ai-first-designAI interaction pattern recommendationsLOW

Gate Enforcement

  • HIGH: Output includes a "Synthesis Judgments Summary" listing each AI judgment call. Acknowledgment prompt required before design recommendations are generated.
  • MEDIUM: Output includes a "Validation Required" section with placeholder for named validation source (expert name, user data reference, or study citation). Design recommendations withheld until validation is provided.
  • LOW: Output template structurally omits the design recommendation section. Tagged with [REFERENCE-ONLY] in title. Notice: "This output reflects AI synthesis from training data. It does not contain design recommendations."

Full protocol documented in skills/user-experience/rules/synthesis-validation.md.

Cross-Framework Synthesis Protocol

When the orchestrator invokes multiple sub-skills for the same engagement (e.g., Heuristic Eval + HEART Metrics for a post-launch UX assessment), it produces a cross-framework synthesis after all sub-skill outputs are available:

Trigger: Two or more sub-skill outputs exist for the same engagement ID.

Mechanism (4-step sequential):

  1. Signal Extraction: The orchestrator reads each sub-skill output's findings/recommendations sections and extracts actionable signals (findings rated severity >= 2 for heuristic eval; metrics below target for HEART; unvalidated assumptions for Lean UX; etc.).
  2. Convergence Detection: Signals from different frameworks that point to the same UX problem are grouped. Convergent signals (2+ frameworks identify the same issue) receive HIGH synthesis confidence. Single-framework signals receive MEDIUM.
  3. Contradiction Identification: Signals from different frameworks that recommend opposing actions are flagged as contradictions. Contradictions always receive LOW synthesis confidence with both positions presented and no resolution attempted — the user decides (P-020).
  4. Unified Output: A synthesis report is produced with three sections: (a) Convergent Findings (HIGH confidence), (b) Single-Framework Findings (MEDIUM), (c) Contradictions (LOW, user decision required). Each finding traces back to its source sub-skill output by engagement ID and finding number.

Output location: projects/${JERRY_PROJECT}/engagements/{engagement-id}/ux-orchestrator-synthesis.md

Failure mode: If synthesis confidence is LOW across > 50% of findings, the orchestrator adds a banner: "Cross-framework synthesis produced mostly low-confidence results. Consider validating individual sub-skill outputs independently before acting on synthesis recommendations." This is a P-022 compliance mechanism — the orchestrator does not overstate the value of synthesis when evidence is weak.

Scope limitation (v1.0.0): Cross-framework synthesis operates on textual output from sub-skills. It does not access MCP design artifacts directly — synthesis inputs are the sub-skill reports, not raw Figma/Miro data. Future versions may add MCP-aware synthesis if usage patterns warrant it.


MCP Integration Architecture

Sub-Skill MCP Dependency Matrix

Sub-SkillFigmaMiroStorybookZeroheightHotjar (Bridge)Whimsical
/ux-heuristic-evalREQ--ENH------
/ux-jtbd--ENH--------
/ux-lean-uxENHREQ----ENH--
/ux-heart-metrics--------ENH--
/ux-atomic-designENH--REQENH----
/ux-inclusive-designREQ--ENH------
/ux-behavior-design--ENH----ENH--
/ux-kano-model--ENH--------
/ux-design-sprintREQREQ------ENH
/ux-ai-first-designREQ--ENH------

REQ = Required (degraded mode + explicit error on failure). ENH = Enhancement (cosmetic limitation on failure).

Figma Dependency Risk Profile

Figma is the highest-risk MCP dependency: 4 sub-skills require it, 2 are enhanced by it (6 of 10 total connections).

Sub-SkillNon-Figma Fallback
/ux-heuristic-evalScreenshot-input mode: user provides design screenshots as image inputs
/ux-design-sprintMiro-only mode: sprint exercises in Miro; manual prototype reference
/ux-inclusive-designScreenshot-input mode: manual component screenshots for evaluation
/ux-ai-first-designManual design description: text-based interaction pattern analysis

Cost Tiers

TierMonthly CostSub-Skills Available
Free$0HEART, JTBD, Kano, Behavior Design (+ Storybook for Atomic Design)
Minimal~$46+ Heuristic Eval, Design Sprint, Lean UX, Inclusive Design, AI-First Design
Full Enhancement~$145-245All 10 with full enhancement MCPs

Current Jerry MCP Integration

MCP ToolUsageAgents
Context7 (resolve-library-id)Resolve UX framework libraries and design system packagesux-atomic-architect, ux-ai-design-guide, ux-sprint-facilitator
Context7 (query-docs)Query component library documentation, accessibility API docsux-atomic-architect, ux-inclusive-evaluator, ux-ai-design-guide, ux-sprint-facilitator

Fallback: Text-Only Mode

When MCP tools are unavailable, all agents operate in text-only mode: users provide design descriptions, screenshots, or markup instead of live design artifacts. The agent methodology remains identical; only the input modality changes. Agents MUST note "text-only mode -- no live design artifact access" in their output.

MCP coordination registry documented in skills/user-experience/rules/mcp-coordination.md.


Cross-Skill Integration

Integration Matrix

Jerry SkillIntegration TypeDirectionDetails
/problem-solvingResearch supportUpstreamps-researcher provides market research for JTBD competitive job analysis; ps-analyst supports Kano survey data interpretation
/adversaryQuality enforcementApplied to outputsS-014 quality scoring on UX deliverables at wave transitions; full adversarial critique for C2+ artifacts per H-13/H-14
/worktrackerOperational trackingInfrastructureTracks UX work items: sub-skill stories, wave transition tasks, Enabler status
/orchestrationWorkflow coordinationCoordinates sub-skillsManages multi-framework workflows; barrier-sync between sub-skills
/nasa-seRequirements handoffDownstreamUX requirements from JTBD job statements and heuristic findings feed into technical requirements and V&V criteria
/diataxisDocumentation methodComplementaryComponent documentation (from Atomic Design) and design system guides use Diataxis methodology
/eng-teamSecurity integrationComplementaryInclusive design findings feed into eng-security accessibility review; component taxonomy informs secure frontend patterns
/astEntity validationInfrastructureWorktracker entities for UX work items validated via AST-based frontmatter extraction (H-33)

Canonical Multi-Skill Workflow Sequences

SequenceSkills InvolvedUse Case
Discovery to Sprint/ux-jtbd then /ux-design-sprintJTBD job statement feeds Design Sprint challenge statement
Sprint to Iterate to Measure/ux-design-sprint then /ux-lean-ux then /ux-heart-metricsSprint produces prototype; Lean UX iterates; HEART measures
Evaluate to Diagnose to Measure/ux-heuristic-eval then /ux-behavior-design then /ux-heart-metricsCrisis mode for urgent UX problems
Build to Evaluate/ux-atomic-design then /ux-inclusive-designBuild components, then evaluate accessibility
Discover to Prioritize/ux-jtbd then /ux-kano-modelDiscover jobs, then prioritize features

Cross-Sub-Skill Handoff Data

Handoffs use the Jerry handoff protocol (schema specified in agent-development-standards.md [Handoff Protocol]; canonical path docs/schemas/handoff-v2.schema.json, pending file creation) with UX-specific artifact types:

FromToHandoff Artifact
/ux-jtbd/ux-design-sprintJob statement + switch forces
/ux-jtbd/ux-kano-modelJob-derived feature list
/ux-design-sprint/ux-lean-uxValidated prototype + Day 4 findings
/ux-heuristic-eval/ux-behavior-designSeverity-rated findings
/ux-atomic-design/ux-inclusive-designComponent inventory with Storybook references
/ux-lean-ux/ux-heart-metricsValidated/invalidated hypothesis backlog

Handoff schema documented in skills/user-experience/rules/ux-routing-rules.md [PARTIAL: EPIC-001 — routing table populated with stage mappings, qualification questions, and common intent resolution; full handoff schema pending].


Constitutional Compliance

All agents adhere to the Jerry Constitution v1.0:

PrincipleRequirementConsequence of Violation
P-003NEVER spawn recursive subagents -- max 1 levelAgent hierarchy violation; uncontrolled token consumption
P-020NEVER override user intent -- ask before destructive opsUnauthorized action; trust erosion
P-022NEVER deceive about actions, capabilities, or confidenceGovernance undermined; quality assessment invalidated
P-001NEVER present findings without evidence or source citationsUnreliable outputs; unfounded claims propagate downstream
P-002NEVER leave outputs in transient context only -- persist to filesContext rot vulnerability; artifacts lost on session compaction
P-004NEVER omit reasoning provenance or source documentationUntraceable decisions; audit trail broken
P-011NEVER make recommendations without supporting evidenceUnsupported recommendations; confidence inflated without basis

Per-agent enforcement: Every agent (orchestrator + 10 sub-skill agents) declares in .governance.yaml:

  • constitution.principles_applied: P-003, P-020, P-022 (minimum; additional principles per agent)
  • capabilities.forbidden_actions: Minimum 3 entries in NPT-009 format referencing the constitutional triplet
  • Sub-skill agents: disallowedTools: [Agent] in .md frontmatter (P-003 enforcement)
  • Domain-specific forbidden actions added per agent role

Quick Reference

Common Workflows

NeedAgentCommand Example
Usability auditux-heuristic-evaluator"Evaluate this dashboard against Nielsen's 10 heuristics"
User motivation researchux-jtbd-analyst"Map the jobs to be done for onboarding"
Experiment designux-lean-ux-facilitator"Create a Lean UX hypothesis for the new search feature"
UX metrics definitionux-heart-analyst"Define HEART metrics for the checkout flow"
Component taxonomyux-atomic-architect"Build an atomic design inventory for the form components"
Accessibility auditux-inclusive-evaluator"Audit this page for WCAG 2.2 AA compliance"
Behavior diagnosisux-behavior-diagnostician"Diagnose why users abandon signup using Fogg B=MAP"
Feature prioritizationux-kano-analyst"Classify these backlog features using Kano analysis"
Rapid prototypingux-sprint-facilitator"Facilitate a design sprint for the mobile navigation redesign"
AI interaction designux-ai-design-guide"Design the conversational UX for the AI assistant feature"
Comprehensive auditux-orchestrator"Full UX audit of the settings experience"
Urgent triageux-orchestrator"CRISIS: users are abandoning checkout"

Agent Selection Hints

KeywordsLikely Agent
heuristic, usability, Nielsen, severity, inspection, evaluation, interface reviewux-heuristic-evaluator
jobs to be done, JTBD, switch interview, outcome, motivation, hiring criteria, user jobsux-jtbd-analyst
lean UX, hypothesis, experiment, MVP, validated learning, assumption, build-measure-learnux-lean-ux-facilitator
HEART, metrics, happiness, engagement, adoption, retention, task success, GSM, measurementux-heart-analyst
atomic design, atoms, molecules, organisms, components, design tokens, design systemux-atomic-architect
accessibility, WCAG, ARIA, screen reader, contrast, cognitive load, inclusive, a11yux-inclusive-evaluator
behavior, Fogg, B=MAP, motivation, ability, prompt, trigger, nudge, behavioral bottleneckux-behavior-diagnostician
Kano, must-be, attractive, one-dimensional, satisfaction, feature classification, delighterux-kano-analyst
design sprint, map, sketch, decide, prototype, test, GV sprint, rapid validationux-sprint-facilitator
AI design, conversational UX, prompt design, trust calibration, AI-first, LLM interface, agenticux-ai-design-guide

Routing Disambiguation

ConditionUse InsteadWhy
Security-focused review or threat modeling/eng-teamUX methodology produces usability findings, not threat models
Adversarial quality review or tournament/adversaryUX agents evaluate design quality, not deliverable quality
General research without UX focus/problem-solving11 UX agents loaded when task requires general methodology
Systems engineering or V&V/nasa-seUX produces user-centered requirements, not system-level V&V
Offensive security testing/red-teamUX methodology produces usability reports, not penetration findings

References

Agent Definition Files

Status key: Exists (stub) = stub file created with frontmatter and minimal sections. [PLANNED: Wave N] = file not yet created; created during that wave's implementation in PROJ-022.

AgentDefinitionGovernanceStatus
ux-orchestratorskills/user-experience/agents/ux-orchestrator.mdskills/user-experience/agents/ux-orchestrator.governance.yamlExists (stub)
ux-heuristic-evaluatorskills/ux-heuristic-eval/agents/ux-heuristic-evaluator.mdskills/ux-heuristic-eval/agents/ux-heuristic-evaluator.governance.yamlExists (stub)
ux-jtbd-analystskills/ux-jtbd/agents/ux-jtbd-analyst.mdskills/ux-jtbd/agents/ux-jtbd-analyst.governance.yamlExists (stub)
ux-lean-ux-facilitatorskills/ux-lean-ux/agents/ux-lean-ux-facilitator.mdskills/ux-lean-ux/agents/ux-lean-ux-facilitator.governance.yaml[PLANNED: Wave 2]
ux-heart-analystskills/ux-heart-metrics/agents/ux-heart-analyst.mdskills/ux-heart-metrics/agents/ux-heart-analyst.governance.yaml[PLANNED: Wave 2]
ux-atomic-architectskills/ux-atomic-design/agents/ux-atomic-architect.mdskills/ux-atomic-design/agents/ux-atomic-architect.governance.yaml[PLANNED: Wave 3]
ux-inclusive-evaluatorskills/ux-inclusive-design/agents/ux-inclusive-evaluator.mdskills/ux-inclusive-design/agents/ux-inclusive-evaluator.governance.yaml[PLANNED: Wave 3]
ux-behavior-diagnosticianskills/ux-behavior-design/agents/ux-behavior-diagnostician.mdskills/ux-behavior-design/agents/ux-behavior-diagnostician.governance.yaml[PLANNED: Wave 4]
ux-kano-analystskills/ux-kano-model/agents/ux-kano-analyst.mdskills/ux-kano-model/agents/ux-kano-analyst.governance.yaml[PLANNED: Wave 4]
ux-sprint-facilitatorskills/ux-design-sprint/agents/ux-sprint-facilitator.mdskills/ux-design-sprint/agents/ux-sprint-facilitator.governance.yaml[PLANNED: Wave 5]
ux-ai-design-guideskills/ux-ai-first-design/agents/ux-ai-design-guide.mdskills/ux-ai-first-design/agents/ux-ai-design-guide.governance.yaml[PLANNED: Wave 5]

Rule Files

Rule files are [STUB: EPIC-001 Foundation] — stub files created during PROJ-022 Foundation phase with section structure and TODO markers. Full implementation in EPIC-001.

Rule FilePurposeStatus
skills/user-experience/rules/ux-routing-rules.mdLifecycle-stage triage logic, cross-sub-skill handoff schema[PARTIAL: EPIC-001]
skills/user-experience/rules/synthesis-validation.md3-tier synthesis hypothesis confidence gates[STUB: EPIC-001]
skills/user-experience/rules/wave-progression.mdWave state tracking and ABANDON log[STUB: EPIC-001]
skills/user-experience/rules/mcp-coordination.mdMCP server coordination registry, maintenance owner[STUB: EPIC-001]
skills/user-experience/rules/ci-checks.mdCI test gate specifications (P-003 enforcement)[STUB: EPIC-001]
skills/user-experience/rules/metrics-plan.mdPost-launch metrics measurement plan[PLANNED: EPIC-008]

Templates

Template files are [STUB: EPIC-001 Foundation] — stub files created during PROJ-022 Foundation phase with template structure. Full implementation in EPIC-001.

TemplatePurposeStatus
skills/user-experience/templates/kickoff-signoff-template.mdKICKOFF-SIGNOFF.md for Wave 0 to 1 transition[STUB: EPIC-001]
skills/user-experience/templates/wave-signoff-template.mdWAVE-N-SIGNOFF.md for wave transitions[STUB: EPIC-001]

Project and Spec Links

ItemLocation
GitHub Issue (authoritative spec)#138
Issue body (main spec)projects/PROJ-020-feature-enhancements/work/issue-drafts/gh-ready/issue-body.md
Tech spec (agents, MCP, cognitive modes)projects/PROJ-020-feature-enhancements/work/issue-drafts/gh-ready/comment-2-tech-spec.md
Appendices (directory structure, scores)projects/PROJ-020-feature-enhancements/work/issue-drafts/gh-ready/comment-3-appendices.md
Acceptance criteriaprojects/PROJ-020-feature-enhancements/work/issue-drafts/gh-ready/comment-1-acceptance-criteria.md
PROJ-022 PLAN.mdprojects/PROJ-022-user-experience-skill/PLAN.md
PROJ-022 WORKTRACKER.mdprojects/PROJ-022-user-experience-skill/WORKTRACKER.md

Standards References

StandardLocation
Agent Definition Format (H-34).context/rules/agent-development-standards.md
Agent Governance Schemadocs/schemas/agent-governance-v1.schema.json
Skill Standards (H-25, H-26).context/rules/skill-standards.md
Quality Enforcement SSOT (H-13, H-14).context/rules/quality-enforcement.md
Mandatory Skill Usage (H-22).context/rules/mandatory-skill-usage.md
Agent Routing Standards (H-36).context/rules/agent-routing-standards.md
Handoff Schemadocs/schemas/handoff-v2.schema.json (canonical path per agent-development-standards.md; file pending creation)
ADR: UX Skill Architectureprojects/PROJ-022-user-experience-skill/decisions/ADR-PROJ022-001-ux-skill-architecture.md (PROVISIONAL)
ADR: Wave Criteria Gatesprojects/PROJ-022-user-experience-skill/decisions/ADR-PROJ022-002-wave-criteria-gates.md (PROVISIONAL)
MCP Tool Standards.context/rules/mcp-tool-standards.md

UX Framework References

FrameworkSourceYearURL
Nielsen's 10 Usability HeuristicsJakob Nielsen, Nielsen Norman Group1994 (updated 2024)https://www.nngroup.com/articles/ten-usability-heuristics/
Jobs-to-Be-DoneClayton Christensen, Anthony Ulwick (ODI)2003/2016https://jobs-to-be-done.com/jobs-to-be-done-a-framework-for-customer-needs-c883cbbbe61f
Lean UXJeff Gothelf and Josh Seiden2013 (3rd ed. 2021)https://www.jeffgothelf.com/lean-ux-book/
Google HEART FrameworkKerry Rodden, Hilary Hutchinson, Xin Fu2010https://research.google/pubs/measuring-the-user-experience-on-a-large-scale-user-centered-metrics-for-web-applications/
Atomic DesignBrad Frost2016https://atomicdesign.bradfrost.com/
WCAG 2.2W3C Web Accessibility Initiative2023https://www.w3.org/TR/WCAG22/
Microsoft Inclusive DesignMicrosoft2016https://inclusive.microsoft.design/
Fogg Behavior ModelBJ Fogg, Stanford Persuasive Tech Lab2009 (B=MAP 2019)https://behaviormodel.org/
Kano ModelNoriaki Kano et al.1984Kano, N., Seraku, N., Takahashi, F., & Tsuji, S. (1984). Attractive Quality and Must-Be Quality. Journal of the Japanese Society for Quality Control, 14(2), 39-48.
AJ&Smart Design Sprint 2.0Jake Knapp / AJ&Smart2016/2019https://ajsmart.com/design-sprint

Research Provenance

ArtifactLocationCreatedQuality Gate
Architecture Visionprojects/PROJ-020-feature-enhancements/work/analysis/ux-skill-architecture-vision.md2026-03-03C3 initial score 0.83, iterated to C4
Framework Selection Analysisprojects/PROJ-020-feature-enhancements/work/analysis/ux-framework-selection.md2026-03-03C3 (part of architecture vision pipeline)
UX Frameworks Surveyprojects/PROJ-020-feature-enhancements/work/research/ux-frameworks-survey.md2026-03-02C3 research (ps-researcher)
Tiny Teams Researchprojects/PROJ-020-feature-enhancements/work/research/tiny-teams-research.md2026-03-02C3 research (ps-researcher)
MCP Design Tools Surveyprojects/PROJ-020-feature-enhancements/work/research/mcp-design-tools-survey.md2026-03-02C3 research (ps-researcher)
Tournament Reports (Iter 1-8)projects/PROJ-020-feature-enhancements/work/issue-drafts/tournament-iter1/ through tournament-iter8/2026-03-03C4 tournament (8 iterations, final spec passed)

Skill Version: 1.0.0 Constitutional Compliance: Jerry Constitution v1.0 SSOT: .context/rules/quality-enforcement.md (thresholds), GitHub Issue #138 (spec) PROJ-022: User Experience Skill Last Updated: 2026-03-03

Keep looking

Skills are one crate of 614,327. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.