bitovi

Orchestrates the full code-to-Figma rebuild workflow for a web application. Thin dispatcher that delegates all substantial work to subagents and reads only small summary files. Runs eight tracked phases (phase0a, phase0b, phase1, phase2, phase2_5, phase3, phase4, phase5). Never calls use_figma or get_screenshot directly.

Subagent for figma-from-code Phase 0a. Discovers the complete component architecture of a web application via browser crawling and static code scanning. Produces a topologically-sorted build order and inspects the target Figma file for existing pages and components.

Standalone validator for a completed figma-from-code rebuild. Navigates the live app, screenshots each component in its natural context, and compares against the matching Figma variant. Also used by the Phase 5 orchestrator agent as a reference for the Component App Map and comparison scripts.

Subagent for figma-from-code Phase 0b. Discovers all Lucide icons and SVG file assets imported across the codebase via static analysis. Extracts SVG markup and maps which components use which icons. No dev server or Figma access needed.

Subagent for figma-from-code Phase 1. Creates Figma variable collections from the project's CSS and Tailwind tokens, extracts the CSS-variable-to-Figma-variable-ID map, and pre-computes all CSS colors as sRGB values.

Subagent for figma-from-code Phase 2. Creates the Figma file page skeleton, foundations documentation frames, and container frames needed by later phases. Requires variable collections to already exist.

Subagent for figma-from-code Phase 2.5. Captures app screenshots and structured text content for every UI component before Figma building begins. Produces a unified batch manifest to minimize browser navigations.

Subagent for figma-from-code Phase 4. Builds full-page screen frames in Figma by composing built component instances into 1440×900 layouts. Validates each screen against app screenshots and iterates up to 3 fix passes.

Always start a fresh browser session after any file change, walk through the full user flow, and monitor for errors before proceeding with further work.

Set up and run different development service arrangements for testing specific features. Use when starting dev servers, testing Storybook addon (SB8 or SB10), testing Tailwind v3 vs v4, testing Astro, or when asked which services to run. Covers compound tasks, port assignments, and which services need each other.

Defines the interaction behavior of global mode buttons (Select, Insert) and component row buttons (Place, Replace). Use when modifying button states, colors, labels, tab visibility, or page interaction behavior. Must review and update these tables before making changes.

Canonical naming for all Convey UI parts, features, and concepts. Reference when discussing, documenting, or writing code for any part of the app. Ensures consistent terminology across humans and AI agents.

Build panel UI components that live-preview Tailwind className changes in the user's app via WebSocket. Use when creating any control (dropdown, scrubber, color picker, toggle, etc.) that lets users hover/scrub values before committing. Covers the preview/revert/stage lifecycle, the onHover/onLeave prop contract, and the focus-trap pattern that prevents preview leaks.

Debug and fix failing Playwright E2E tests. Use when tests fail, when asked to fix failing tests, or when investigating test failures. Analyzes test output, screenshots, error context, and uses Playwright MCP to identify root causes.

Canonical terminology for Tailwind CSS concepts used throughout the codebase. Reference this skill to ensure consistent naming in code, docs, and conversation.

Create React components or hooks following the modlet pattern in this project. Use when creating any new component or hook in panel/src/components/. Modlets are self-contained folders with index.ts, implementation, tests, and optional types.

Sub-skill: Analyze a single Figma design frame via the figma-frame-data MCP tool. Use this variant when local files are NOT available (e.g., GitHub cloud Copilot where curl is blocked). Calls figma-frame-data with a batchToken to retrieve image, structure, and context via MCP, then analyzes and returns the analysis as text. No filesystem reads needed — pure MCP.

Sub-skill: Fetch raw content for one or more URLs via the extract-linked-resources MCP tool. Supports Jira issues, Confluence pages, Google Docs, and Google Sheets. Saves content to .temp/cascade/context/ and appends newly discovered links to to-load.md. Used as a building block by parent skills like generate-questions and write-jira-story.

Generate frame-specific clarifying questions about ambiguous UI behaviors from a Jira epic and its linked Figma designs, Confluence pages, and Google Docs. Uses iterative content loading, parallel Figma frame analysis, and cross-content synthesis to produce targeted behavior questions organized by Figma frame.

Write or refine a Jira story description with full context from Figma designs, Confluence docs, Google Docs, and parent epic. Gathers all linked resources, analyzes Figma frames, runs scope analysis, and writes a comprehensive story with User Story Statement, Scope Analysis, Acceptance Criteria (Gherkin), NFRs, and Developer Notes. Uses ☐/✅/❌/❓/💬 scope markers and flips ❓→💬 when answers are found.

Write a full Jira story (User Story Statement, Scope Analysis, Acceptance Criteria in Gherkin, NFRs, Developer Notes) from the next unwritten shell story in a Jira epic. Loads only the Figma screens listed in that shell story, runs scope analysis anchored to its scope bullets, generates the full story description, creates a Jira story under the epic, adds blocker links for dependencies, and marks the shell story complete in the epic.

Write or refresh the Shell Stories section of a Jira epic by loading all linked context (Figma, Confluence, Google Docs), analyzing every Figma frame in parallel, running scope analysis, then generating a prioritized list of incremental shell story outlines grouped by user workflow. Preserves completion markers for already-written stories. Uses ☐/⏬/❌/❓ scope markers with SCREENS and DEPENDENCIES per story.

Sub-skill: Produce a Scope Analysis from frame analyses and all gathered context. This is the critical step that takes per-frame analyses + epic context + Confluence/Google Docs + Figma comments and categorizes every feature by scope (☐/✅/⏬/❌/❓/💬). Groups features by user workflow, not by screen. Supports self-healing ❓→💬 flipping on re-runs. Output drives all downstream work (questions, shell stories, story writing).

Sub-skill: Analyze a single Figma design frame from local files. Reads image.png (vision), structure.xml (component tree), and context.md (comments/annotations) from .temp/cascade/figma/{fileKey}/frames/{name}/. Writes analysis.md. Designed to run as a subagent — no MCP tools needed, pure filesystem.

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies

Use when you have a written implementation plan to execute in a separate session with review checkpoints

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup

Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation

Use when completing tasks, implementing major features, or before merging to verify work meets requirements

Use when executing implementation plans with independent tasks in the current session

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

Initialize a minimal hello-world mock-app with React, Vite, TypeScript, Tailwind CSS, React Router, and MSW. Use when scaffolding a new mock-app from scratch, bootstrapping a project skeleton, or starting a fresh prototype. Produces a working app with routing, mock API support, and a hello-world landing page.

Track and manage reusable UI components and patterns in the mock-app. Maintains an inventory of extracted components and unextracted patterns to promote incremental refactoring and prevent code duplication.

End-to-end workflow for implementing a new feature in the mock-app, ensuring data model alignment, sample data, UI/UX consistency, component reuse, and automated verification with Playwright MCP.

Create new Agent Skills for this project. Use when asked to create a skill, document a workflow, or teach Copilot a new capability. Skills are stored in .github/skills/ and can include instructions, scripts, examples, and resources.

Debug and fix failing Playwright E2E tests. Use when tests fail, when asked to fix failing tests, or when investigating test failures. Analyzes test output, screenshots, error context, and uses Playwright MCP to identify root causes.

Always start a fresh browser session after any file change, walk through the full user flow, and monitor for errors before proceeding with further work.

Create comprehensive feature requirement documents for the WCU Mosquito Surveillance Web Application. Use when asked to document a feature, write feature requirements, or create feature specifications from the roadmap. Includes user stories, functional requirements, data model review, workflows, and edge cases.

Extract reusable UI components from inline patterns. Covers component design, TypeScript props, Storybook stories, refactoring strategy, and best practices for creating shared UI primitives.

Extract and compare computed CSS styles between a baseline URL and a dev/Storybook URL using Playwright MCP evaluate calls. Use when visual differences have been identified and you need to find the exact CSS property mismatches to fix. Pairs with the visual-diff skill for a complete visual QA workflow.

Debug and fix failing Playwright E2E tests. Use when tests fail, when asked to fix failing tests, or when investigating test failures. Analyzes test output, screenshots, error context, and uses Playwright MCP to identify root causes.

Diagnose and fix UI/runtime bugs. After a code change, verify in the browser with the available browser tools, read the console, and — for non-obvious bugs — add strategic temporary logging to locate the cause, then remove it. The agent decides how much of this flow a given bug needs.

Discover all visual states of a component by interactively exploring the production baseline page — not by reading story files. Finds every interactive element (nav items, dropdowns, toggles, buttons) that produces a distinct visual state, records the interactions needed to reach each one, and saves a pixel-perfect.config.json that the pixel-perfect skill uses to replay those same interactions on both production and Storybook. Run this once per component; re-run when the production page changes.

End-to-end workflow for implementing a new UI feature in a mock app. Ensures data model alignment, sample data, UI/UX consistency, component reuse, and automated verification. Load for ANY UI change.

Orchestrate a complete visual QA workflow to make a dev/Storybook page match a baseline URL pixel-for-pixel. Combines the visual-diff skill (screenshot comparison) with the computed-styles skill (CSS property extraction) in a loop until the pages match. Use when replicating a production page, matching a design reference, or achieving pixel perfection on a component.

Redeploy the "Measure of Success" Jira Forge app to a dev site after code or manifest changes. Use when asked to deploy, redeploy, push to Jira, install/upgrade the app, ship a new module/scope/custom field, or start a live tunnel. Covers Node 22, the Custom UI build, forge deploy/install --upgrade, and the tunnel.

Make UIs responsive across all devices and verify responsiveness using Playwright. Includes scrollbar testing, breakpoint verification, and layout validation. Use when implementing new features, fixing layout issues, or ensuring mobile compatibility.

Post generated design behavior questions as comments on Figma frames. Takes questions organized by frame (from generate-behavior-questions skill) and posts each question as a comment pinned to the correct Figma frame using the figma-post-comment MCP tool.

Generate frame-specific clarifying questions about ambiguous UI behaviors from a Jira epic and its linked Figma designs, Confluence pages, and Google Docs. Uses iterative content loading, parallel Figma frame analysis, and cross-content synthesis to produce targeted behavior questions organized by Figma frame.

Create new Agent Skills for this project. Use when asked to create a skill, document a workflow, or teach the agent a new capability. Skills are stored in .claude/skills/ and can include instructions, scripts, examples, and resources.

Ensure existing UI components are reused before creating new ones. Use when implementing any UI from Figma designs, tickets, or mockups. Requires a component audit to search the codebase before creating new components. If a component doesn't exist, delegates to figma-implement-component skill.

Manage TypeScript type flow between @carton/shared, @carton/server, and @carton/client packages. Use when adding Prisma models, creating tRPC procedures, importing types in components, or resolving circular dependency issues between packages.

Check a React component against its Figma design source and identify differences. Use when reviewing component implementations, syncing designs, auditing visual accuracy, or updating components to match new Figma designs.

Connect shadcn/ui components to their Figma design system counterparts using Code Connect.

Explore a Figma file to discover all pages, components, and component sets using the Figma REST API. Use when you need to list components in a Figma file, find component node IDs for Code Connect, or understand the structure of a design file.

Final step in building instruction context for codebase

Determine the type of project and summarize the tech stack. Your summary should include:

Second step in building instruction context for codebase

Third step in building instruction context for codebase

Fourth step in building instruction context for codebase

Fifth step in building instruction context for codebase

Connect shadcn/ui components to their Figma design system counterparts using Code Connect.

Explore a Figma file to discover all pages, components, and component sets using the Figma REST API. Use when you need to list components in a Figma file, find component node IDs for Code Connect, or understand the structure of a design file.

Implement React components from Figma designs. Use after figma-design-react has analyzed the design, or directly with a Figma URL to run analysis and implementation in one shot. Delegates to create-react-modlet for folder structure, then adds Figma-specific implementation, stories for each variant, and README with design context. Also triggers on phrases like "build this component from Figma", "implement this design", "create a component from this Figma URL", "turn this Figma into code".

Generates Figma Code Connect mappings for React or Angular repos, validates them with pre-checks and the Figma CLI, and can publish via figma connect. Use when a user provides a Figma file URL and a component repo and asks for Code Connect files, figma.config.json, or validation.

Design React components from Figma files. Use when given a Figma URL to analyze a design and propose React component architecture, props API, and variant handling. Outputs design analysis and suggested API, does not build components.

How to manage deployments of the FreshBooks MCP Server to AWS EC2