By thekostakis
Visual design system toolkit with UX/usability expert review. Consultant skill for establishing design systems via interview or site extraction with page discovery, parallel fingerprinting, fluid responsive detection, and custom component motion discovery; codebase reverse-engineering mode derives tokens and component specs from the repo with chrome-devtools-mcp route inspection (spec at docs/design-system/codebase-reverse-spec-*.md, section-by-section approval, per-file delete confirmation). Component-context skill plus opus component-context agent: minimal return (one full spec + motion) for exact or high-confidence fuzzy; auto top-3 full specs + pattern synthesis + motion when ambiguous; gap inference when no match — never the whole compendium. Design-reviewer skill and agent: headless Playwright quality gate (playwright-skill-bridge.mjs) — CI-safe, no DevTools MCP — Nielsen heuristics UX scoring (0-100), diff mode, review-checklist synergy, functional/behavior-change escalation, Categories A–F in phase files. Opus + haiku mechanical dispatch; .agent-progress/ logs. Update-component-index skill for the component compendium index.
Based on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Use when the user needs design-system guidance for UI implementation without loading the entire compendium. Triggers: which component to use, pick/recommend a component, component mapping, "no spec for", missing component spec, how to design or style a widget, compose components, UI pattern → design system, component gap, fuzzy matches. Behavior: **minimal** output when exact or high-confidence fuzzy: the **entire** matched component spec file (full markdown body) + motion — not a digest; **top-3** full specs + pattern synthesis + motion when fuzzy is ambiguous; **gap** mode when nothing scores. Always include **motion guidance** (from specs + design-guidelines.md). Not for visual-design-consultant (new system) or design-reviewer (QA).
Use when a visual/UI implementation step has been completed and needs design system review, or when reviewing components against requirements with visual output. Triggers on: component implementation complete, page complete, design review needed, visual quality gate, epic/milestone with UI work. Do NOT use for backend-only or CLI work with no visual output.
This skill should be used for design-system guidance during UI implementation: which component to use, specs for named components, fuzzy matches when no exact spec exists, or how to approach a component that has no spec yet — without loading the entire compendium into the caller's context. Trigger phrases: "load component spec", "component design spec", "what does the design say about [component]", "design spec for [component]", "which component should I use", "recommend a component", "pick a component from the design system", "map this UI to components", "no spec for", "missing component spec", "there's no design for", "how should I design", "how to style", "design guidance for", "component gap", "compose from existing components", "what should I use for [pattern]", "implement this against the design system", "offload component context", "get a compact spec briefing". Also when creating or editing files under components/, ui/, or widgets/ where token-aware implementation hints are needed. Prefer dispatching the visual-design:component-context agent. **Exact match or ~90%+ high-confidence fuzzy:** return the **complete** markdown body of the matched component spec (entire file), plus motion guidance — never a summary instead of the full spec. **Top-3** when fuzzy is ambiguous; **gap** when nothing fits. Never dump the whole compendium. Do NOT use for establishing a new design system from scratch (visual-design-consultant) or for post-build visual QA (design-reviewer).
This skill should be used when a major project step involving visual/UI work has been completed and needs to be reviewed against the design system. Examples: "I've finished implementing the dashboard page", "the card component is done", "implement epic 3" (when epic involves visual output). Also fires on: "review component", "design review", "check against design system", "visual review", "run design tests", "implement epic", "implement milestone", "implement feature", "perform a design review of [requirements]". Do NOT trigger for backend-only, API-only, or CLI work with no visual output. Uses **headless Playwright** (`playwright-skill-bridge.mjs`) for inspection — AFK-safe and CI-safe; no Chrome DevTools MCP. Includes Nielsen's 10 Usability Heuristics with UX scoring (0-100) and diff mode for follow-up reviews. When dispatched as the design-reviewer agent, writes granular progress to `.agent-progress/` (see `references/agent-progress.md`) and emits short parent summaries for orchestrators.
This skill should be used to regenerate the component index after manually adding, editing, or removing component spec files. Trigger phrases: "update component index", "rebuild component index", "refresh design index", "regenerate component index".
This skill should be used when the user wants to establish, extract, or update a visual design system. Trigger phrases: "design system", "visual design", "design guidelines", "extract design from site", "design tokens", "component styles", "I want it to look like [site]", "set up the design", "design interview", "update design guidelines", "add to design system". Also trigger when the user provides a URL and asks about its design, or when they describe visual preferences for a project. Codebase reverse-engineering triggers: "design system from codebase", "reverse engineer components from the repo", "extract tokens from this project", "derive design guidelines from code", "build component specs from existing UI", "motion design from codebase". Do NOT trigger for requirements gathering — that is the requirements-gatherer skill. Do NOT trigger for code review — that is the design-reviewer skill.
Uses power tools
Uses Bash, Write, or Edit tools
A Claude Code plugin marketplace for product design, project planning, and testing workflows.
A senior product consultant that interviews you to understand what you want to build, then produces a structured requirements.md file. It defines WHAT and WHY, never HOW — no architecture, no tech recommendations, no implementation details.
Supports two modes:
New mode — when no requirements.md exists, runs a full requirements interview from scratch.
Addendum mode — when requirements.md already exists, reads it first, then interviews you about what's new or changing. Produces a separate requirements-addendum-[date].md file. The original is never modified.
Trigger phrases:
What it does:
Reads a reviewed requirements.md or requirements-addendum-*.md and creates structured epics and issues in GitHub or Jira. You trigger this after reviewing and editing the requirements document.
Supports two modes:
Full mode — reads requirements.md, groups features into dependency-ordered epics, creates milestones, labels, and issues from scratch.
Addendum mode — reads a requirements-addendum-*.md file, fetches existing milestones and issues, then creates only what's new or changed:
needs-decision issuesTrigger phrases: "create issues from requirements", "organize into epics", "push to GitHub", "push to Jira", "create tickets"
What it does:
backlog.md file if no GitHub/Jira integration is availableneeds-decision-labeled issuesEstablishes a project's visual design system through plain-language interview or by extracting patterns from example websites. Produces design-guidelines.md (core tokens, permanently in CLAUDE.md context) and a component compendium (design/components/) with detailed per-component specs.
Trigger phrases: "design system", "visual design", "extract design from [site]", "I want it to look like [site]", "design guidelines"
What it does:
visual-design:component-context agent)Off-thread lookup with three behaviors: (1) exact or high-confidence (~90%) fuzzy → the entire matched component spec file (full markdown) + motion guidance, no long briefing; (2) ambiguous fuzzy → auto top 3 full specs + pattern synthesis + motion; (3) no match → inferred guidance + motion from guidelines. Further candidates stay one-line only — never the whole compendium.
Trigger phrases: see the skill description in the plugin (e.g. "load component spec", "which component should I use", "no spec for", "map this UI to components"). Prefer dispatching the agent when implementing so the main session stays lean.
Senior creative director quality gate. Verifies implemented UI against the design system through chrome-devtools-mcp (live browser inspection): visual appearance, CSS/token compliance, accessibility (axe), motion, responsive behavior, and Nielsen’s 10 usability heuristics with a 0–100 UX score. Supports diff mode (compare to a prior design-review-*.md report) and can consume design/review-checklist.md from the consultant.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimnpx claudepluginhub thekostakis/requirements-gatherer --plugin visual-designRequirements gathering interview and issue creation workflow. Includes skills: a senior product consultant that interviews users to produce structured requirements; source-sync mode to reverse-engineer or reconcile requirements.md from repo, docs, URLs, and attachments (with drift analysis and explicit approval before overwriting); and an organizer that converts reviewed requirements into GitHub milestones/issues or Jira epics/stories.
Playwright functional test generator with full-stack performance analysis. Headless Playwright only (playwright-skill-bridge.mjs + test suite) — CI-safe, AFK-safe, no Chrome DevTools MCP. Playwright AI agents (Planner, Generator, Healer) for resilient test authoring, accessibility-tree selectors, @axe-core/playwright for WCAG audits, visual regression via toHaveScreenshot, Lighthouse CLI / Lighthouse CI for performance and audit scores. TDD-style fix loop, then Lighthouse + axe + performance deep-dive (report-only with escalation tags). Opus agent with haiku TDD sub-agent; .agent-progress/ logs (optional progress_log_path).
Defect reporting and submission workflow with dispatch contract support. Reporter skill conducts structured interviews, auto-generates reproduction steps via headless Playwright (visual bugs; plugin bridge script, CI-safe) or code tracing (API bugs), cross-references requirements, and classifies as true defect, story update, or feature request. Supports dispatch from orchestrators with page URL (visual) or API endpoint identity (non-visual). Organizer skill submits to GitHub, Jira, Linear, or GitLab with system-native formatting and requirement traceability.
Design workflow for coding agents: discovery, research, strategy, inclusive design, critique, and handoff — enforced automatically
Accelerate design workflows — critique, design system management, UX writing, accessibility audits, research synthesis, and dev handoff. From exploration to pixel-perfect specs.
Refine project plans through in-depth Socratic questioning. Interview about technical implementation, UI/UX, concerns, and tradeoffs to produce comprehensive specs.
Streamline design operations with critique frameworks, handoff specs, sprint planning, review processes, and team workflows.
Build, document, and maintain scalable design systems — from tokens and components to accessibility and theming.
Design with Intent. A comprehensive UX and design strategy system — 16 specialized skills and 6 named agents covering strategy, research, experience design, quality assurance, accessibility, measurement, and engineering handoff.