Agent Hub

Explores user intent, requirements and design before any creative work — creating features, building components, adding functionality, or modifying behavior.

Creates detailed implementation plans when you have a spec or requirements for a multi-step task, before touching code.

Executes a written implementation plan in a separate session with review checkpoints.

Executes plans by dispatching fresh subagent per task with two-stage review: spec compliance then code quality.

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

Guides completion of development work by presenting structured options for merge, PR, or cleanup when implementation is complete and tests pass.

Ensures an isolated workspace exists via native tools or git worktree fallback when starting feature work.

Disciplined diagnosis loop for hard bugs and performance regressions. Reproduce → minimise → hypothesise → instrument → fix → regression-test.

Systematic approach to debugging — use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

Test-driven development with red-green-refactor loop for building features or fixing bugs.

Use when implementing any feature or bugfix, before writing implementation code.

Requires running verification commands and confirming output before making any success claims. Evidence before assertions always.

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

Use when receiving code review feedback, before implementing suggestions. Requires technical rigor and verification.

Finds deepening opportunities in a codebase, informed by domain language in CONTEXT.md and decisions in docs/adr/.

Provides broader context or higher-level perspective on a section of code and how it fits into the bigger picture.

Sets up agent skills block in AGENTS.md/CLAUDE.md and docs/agents/ with issue tracker, triage labels, and domain doc layout.

Breaks a plan, spec, or PRD into independently-grabbable issues using tracer-bullet vertical slices.

Turns the current conversation context into a PRD and publishes it to the project issue tracker.

Triages issues through a state machine driven by triage roles — create, review, prepare for AFK agent, manage workflow.

Interviews the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree.

Grilling session that challenges your plan against existing domain model, sharpens terminology, and updates documentation inline.

Compacts the current conversation into a handoff document for another agent to pick up.

Builds a throwaway prototype to flesh out a design — terminal app for logic questions or UI variations toggleable from one route.

Teaches the user a new skill or concept within the current workspace.

Establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions.

Helps users discover and install agent skills when they ask how to do something or want to extend capabilities.

Creates new skills, modifies and improves existing skills, measures skill performance, runs evals, and benchmarks.

Creates new agent skills with proper structure, progressive disclosure, and bundled resources.

Use when creating new skills, editing existing skills, or verifying skills work before deployment.

Template for creating new skills. Replace with description of the skill and when Claude should use it.

Structured workflow for co-authoring documentation — proposals, technical specs, decision docs. Helps transfer context and refine through iteration.

Reviews docs/prose for Writing Guidelines compliance — voice, tone, clarity, and consistency.

Resources for writing internal communications using company formats — status reports, leadership updates, newsletters, FAQs, incident reports.

Create, read, edit, or manipulate Word documents (.docx) — reports, memos, letters, templates with professional formatting.

Everything PDF — reading, extracting text/tables, merging, splitting, rotating, watermarking, creating, form filling, encrypting, OCR.

Create, read, edit PowerPoint presentations (.pptx) — slide decks, pitch decks, templates, layouts, speaker notes.

Open, read, edit, create spreadsheets — .xlsx, .xlsm, .csv, .tsv. Formulas, formatting, charts, data cleaning, conversions.

Toolkit for styling artifacts with a theme — slides, docs, reports, HTML pages. 10 pre-set themes or generate new ones on-the-fly.

Applies Anthropic's official brand colors and typography to any artifact needing the company look-and-feel.

Guidance for distinctive, intentional visual design when building new UI or reshaping existing one — aesthetic direction, typography, anti-template choices.

Anti-slop frontend skill for landing pages, portfolios, and redesigns. Reads the brief, infers design direction, ships interfaces that don't look templated.

Original v1 taste-skill preserved for projects depending on its exact behavior. Current default is design-taste-frontend (v2).

Teaches the AI to design like a high-end agency. Exact fonts, spacing, shadows, card structures, and animations that make a website feel expensive.

Elite UX/UI & Advanced GSAP Motion Engineer. Python-driven true randomization, strict AIDA page structure, wide editorial typography, gapless bento grids.

Clean editorial-style interfaces. Warm monochrome palette, typographic contrast, flat bento grids, muted pastels. No gradients, no heavy shadows.

Raw mechanical interfaces fusing Swiss typographic print with military terminal aesthetics. Rigid grids, extreme type scale contrast, utilitarian color.

Elite frontend image-direction skill for premium, conversion-aware website design references. ONE separate horizontal image FOR EVERY section.

Elite mobile app image-generation for premium, app-native screen concepts — iOS, Android, cross-platform. Clean hierarchy, tasteful iconography, phone mockup framing.

Elite website image-to-code skill. Generates design images first, analyzes them, then implements website to match. Prefers large, readable, section-specific images.

Upgrades existing websites and apps to premium quality. Audits current design, identifies generic AI patterns, applies high-end standards without breaking functionality.

Creates beautiful visual art in .png and .pdf documents using design philosophy — posters, art, static pieces. Original work, never copies existing artists.

Creates algorithmic art using p5.js with seeded randomness and interactive parameter exploration — generative art, flow fields, particle systems.

Premium brand-kit image generation for high-end brand-guidelines boards, logo systems, identity decks, and visual-world presentations.

Knowledge and utilities for creating animated GIFs optimized for Slack — constraints, validation tools, and animation concepts.

Design mobile apps, create screens, build UI, or interact with Sleek projects — high-level requests and specific ones.

Semantic Design System Skill for Google Stitch. Generates agent-friendly DESIGN.md files enforcing premium, anti-generic UI standards.

Manages shadcn components and projects — adding, searching, fixing, debugging, styling, and composing UI with component registries and presets.

React and Next.js performance optimization guidelines from Vercel Engineering for components, pages, data fetching, and bundle optimization.

React composition patterns that scale — compound components, render props, context providers, component architecture. Includes React 19 API changes.

Guide for implementing smooth animations using React's View Transition API — page transitions, route changes, shared element animations, list reorder.

React Native and Expo best practices for building performant mobile apps — components, list performance, animations, native modules.

Vercel cost and performance optimization for deployed projects — bill reduction, slow routes, caching, Function Invocations, Build Minutes, Core Web Vitals.

Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using React, Tailwind CSS, shadcn/ui — state management, routing.

Toolkit for interacting with and testing local web applications using Playwright — verifying frontend functionality, debugging UI, screenshots, browser logs.

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly.

Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy.

Smart router across full image-model catalog on RunComfy — FLUX 2, Nano Banana 2/Pro, GPT Image 2, Seedream 5/4-5/4-0, Qwen Image, Z-Image Turbo, Wan 2-7.

Smart router for image editing — Nano Banana Edit, GPT Image 2 Edit, Flux Kontext Pro, Z-Image Turbo Inpaint. Swaps backgrounds, removes objects, rewrites text.

Mask-driven image inpainting — object removal, watermark removal, region replacement, blemish cleanup. Routes to Z-Image Turbo Inpainting and identity-preserving edit models.

Extend still images beyond original canvas — change aspect ratio, uncrop, fill surroundings while preserving original content.

Relight a still image — change lighting setup, color temperature, direction, or mood. Product relighting, portrait mood shift, color-grade change.

Generate images with Flux 2 Klein (Black Forest Labs' distilled fast variant) — sub-second latency, multi-reference brand styling, declarative prompts.

Edit images with Flux 1 Kontext Pro — single-reference precise local edits, strong prompt control, consistent high-fidelity outputs.

Generate and edit images with OpenAI GPT Image 2 — embedded text, logos, multilingual typography, instruction precision. 3 fixed sizes.

Edit images with GPT Image 2 edit endpoint — preservation language, in-image text editing, multi-reference up to 10 images, layout precision.

Generate images with Google Nano Banana 2 (Gemini flash-tier) — rapid iteration, in-image typography, predictable framing, optional web-grounded context.

Edit images with Google Nano Banana 2 edit endpoint — preserve subject identity, swap background, spatial language edits, batch up to 20 images.

Smart router across full video-model catalog — HappyHorse 1.0, Wan 2-7, Seedance v2, Kling 3.0/2-6, Veo 3-1, Hailuo 2-3, Dreamina 3-0. Text-to-video and image-to-video.

Animate any still image — HappyHorse 1.0 I2V, Wan 2.7 with audio_url lip-sync, or Seedance 2.0 Pro multi-modal animation.

Edit existing video — Wan 2.7 Edit-Video (restyle/background swap), Kling 2.6 Pro Motion Control (motion transfer), Lucy Edit Restyle (outfit swap).

Extend or continue existing video clips — Google Veo 3-1 extend-video. Consistent motion, lighting, and subject identity across continuation.

Region edits across video frames — remove objects, clean up wires/watermarks, replace regions with matching motion.

Extend spatial canvas of video — change aspect ratio (vertical to horizontal), add environment beyond original frame.

Smart router across music-model catalog — ElevenLabs AI Music (premium 44.1 kHz, $0.0083/s) and ACE Step (open-weights, $0.0002-0.0003/s, ~27× cheaper). Inpaint and outpaint audio.

Generate full songs with ElevenLabs Music — studio-quality 44.1 kHz stereo, 5s-5min, section-level control (Intro/Verse/Chorus/Bridge), multilingual vocals.

Generate, inpaint, and outpaint music with ACE Step (StepFun-AI open-weights) — tag-driven composition, multilingual lyrics, 5s-4min, $0.0002-0.0003/s.

Generate text-to-video with HappyHorse 1.0 — #1 on Arena, native 1080p with in-pass synchronized audio, multi-shot character consistency.

Generate text-to-video with Wan 2.7 (Wan-AI flagship motion model) — multi-reference conditioning, audio-driven lip-sync, smooth transitions.

Generate cinematic short-form video with ByteDance Seedance 2.0 Pro — multi-modal references (up to 9 images, 3 videos, 3 audio), natural lip-sync.

Kling 3.0 video generation — Kuaishou's third-gen multi-shot model with native synchronized audio and consistent character identity. Standard, Pro, and 4K tiers.

Create AI avatar, talking-head, and lip-sync videos — OmniHuman, Wan 2-7 lip-sync, HappyHorse 1.0 with audio, Seedance v2 Pro. Routes based on intent.

Lip-sync a face to audio — OmniHuman (full-body avatar), Sync Labs sync v2/Pro (mouth sync on video), Kling lipsync, Creatify lipsync.

Swap a face/character into video or images — Wan 2-2 Animate, GPT Image 2 Edit, Nano Banana Edit, Flux Kontext, Kling 2-6 Motion Control Pro.

Pose-conditioned generation — Kling 2-6 Motion Control, Wan 2-2 Animate (audio-driven + pose), Z-Image Turbo ControlNet LoRA (OpenPose/DWPose/canny/depth).

Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image. 1536x1872 atlas, 9 animation states.

Run any model on RunComfy from CLI — one binary, one auth, hundreds of endpoints. Image/video gen, edit, lip-sync, face swap, inpainting, upscale, LoRA training.

Deploy applications and websites to Vercel — push live, create preview deployments, manage projects.

Deploy and manage projects on Vercel using token-based authentication — no interactive login needed.

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools.

Edit opencode's configuration — opencode.json, .opencode/, agents, subagents, skills, plugins, MCP servers, permission rules.

Reference for Claude API / Anthropic SDK — model IDs, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration.

Review UI code for Web Interface Guidelines compliance — accessibility, UX, design best practices.