The skills you have.
The one you need.

Migrate C++/CUDA PyTorch extension code to use the PyTorch stable ABI. Use when the user wants to make their project ABI stable, migrate to stable ABI, or use torch::stable APIs.

12-agent academic paper writing pipeline. 11 modes (full/plan/outline/revision/revision-coach/abstract/lit-review/format-convert/citation-check/disclosure/rebuttal-audit). 6 paper types, 5 citation formats, bilingual abstracts, LaTeX/DOCX-via-Pandoc/PDF output. Style Calibration + Writing Quality Check + Anti-Patterns with IRON RULE markers. Triggers: write paper, academic paper, guide my paper, parse reviews, audit my rebuttal, check my response draft, AI disclosure, 寫論文, 學術論文, 引導我寫論文, 審查意見, 評估回覆, 논문 작성, 초록 작성, 논문 수정, 논문 계획을 도와줘, 심사 의견 반영, 답변서 점검, AI 사용 고지.

Multi-perspective academic paper review with dynamic reviewer personas. Runs a 5-seat, role-separated review panel (Journal-Fit Reviewer + 3 peer-review roles + Devil's Advocate) with field-specific expertise; role separation is not a claim of independent error processes. Supports full review, re-review (verification), quick assessment, methodology focus, Socratic guided, and calibration modes. Triggers on: review paper, peer review, manuscript review, referee report, review my paper, critique paper, simulate review, editorial review, calibrate reviewer, reviewer calibration, measure reviewer accuracy, 審查論文, 論文審查, 模擬審查, 同儕審查, 幫我審這篇, 以審查人角度評估, 審查者校準, 논문 심사, 동료 심사, 모의 심사, 심사자 관점에서 평가, 심사자 보정.

Orchestrator for the full academic research pipeline: research -> write -> integrity check -> review -> revise -> re-review -> re-revise -> final integrity check -> finalize. Coordinates deep-research, academic-paper, and academic-paper-reviewer into a seamless 10-stage workflow with mandatory, coverage-bounded integrity checks, two-stage peer review, and auditable quality-assurance artifacts. Triggers on: academic pipeline, research to paper, full paper workflow, paper pipeline, end-to-end paper, research-to-publication, complete paper workflow, 연구부터 논문까지, 연구 주제 설정부터 논문 완성까지, 논문 전체 워크플로.

Access and test Vercel deployments protected by Vercel Authentication, SSO, or Deployment Protection. Use when curl, agent-browser, Playwright, or another automated request reaches a Vercel login or protection page; when a protected preview or production URL returns 401 or 403; when TRUSTED_SOURCES_ENVIRONMENT_MISMATCH appears; or when choosing between `vercel curl` and the `x-vercel-trusted-oidc-idp-token` header.

A dummy skill that announces the arrival of agents. Use when the user asks about agents, the agent revolution, or wants to hype up agents.

Vercel AI Gateway guidance for setup, model discovery, authentication, routing, fallbacks, BYOK, budgets, spend reporting, observability, compatible APIs, and coding-agent configuration. Use when adding AI Gateway to an app, migrating provider calls, choosing models or providers, debugging gateway requests, or running `vercel ai-gateway` commands.

Vercel AI SDK expert guidance. Use when building AI-powered features — chat interfaces, text generation, structured output, tool calling, agents, MCP integration, streaming, embeddings, reranking, image generation, or working with any LLM provider.

Analyze PyTorch CUDA memory snapshots from torch.cuda.memory._dump_snapshot(), torch.cuda.memory._snapshot(), or memory_viz artifacts. Use when the user wants to understand GPU memory usage, top allocation sites, deallocation timing, OOMs, allocator fragmentation, inactive cached memory, memory leaks, retained tensors, or Python reference cycles in PyTorch workloads. Supports diffing two snapshots from the same run to pinpoint growing allocation sites.

Analyze ML experiment results, compute statistics, generate comparison tables and insights. Use when user says "analyze results", "compare", or needs to interpret experimental data.

Reverse-lookup glossary that turns a vague description of a web animation or motion effect into its exact term ("the bouncy thing when a popover opens" → Pop in; "the iOS rubber-band scroll" → Rubber-banding). Use when the user asks "what's it called when…", or describes a motion effect without knowing its name and wants the right word to prompt an AI or designer with. For naming an effect, not designing or building one.

App Store optimization and marketing skills for descriptions, screenshots, keywords, review responses, and comprehensive promotional strategy. Use when user needs help with App Store presence, ASO, marketing, or customer communication.

Apple's approach to interface design and fluid, physical motion, translated for the web. Use when building or reviewing gesture-driven UI, spring animations, drag/swipe/sheet interactions, momentum and interruptible transitions, translucent materials and depth, typography (optical sizing, tracking, leading), reduced-motion, or the design foundations (feedback, spatial consistency, restraint) behind Apple-style interfaces.

Apple Intelligence skills for on-device AI features including Foundation Models, Visual Intelligence, App Intents, and intelligent assistants. Use when implementing AI-powered features.

Autonomously improve a real artifact (code, training recipe, agent harness, data pipeline, prompt) against an objective and an evaluator, using Hypothesis Tree Refinement (HTR) from the Arbor paper. Use this whenever someone wants to iteratively optimize something over many experiments without overfitting — e.g. "get my model's eval score up", "improve this agent/harness", "tune this pipeline", "beat the baseline on this benchmark", "run a search over approaches and keep the best", "do an MLE-bench / Kaggle-style optimization", or any long-horizon "make this artifact better and don't just memorize the dev set" task. Trigger it even when the user doesn't say "Arbor" or "hypothesis tree" but describes repeated experiment-and-evaluate loops, branching exploration of competing ideas, or worries about a dev/test gap. Runs Claude itself as the coordinator with subagent executors in isolated git worktrees; for the standalone `arbor` CLI tool see references/arbor-upstream.md.

Ask clarifying questions before executing tasks. Use this skill whenever the user says "ask me questions first", "clarify before doing", "ask before you start", invokes /ask, or wants Claude to gather requirements through Q&A before taking action. Also trigger when the user provides a task prefixed with "ask:" or says things like "I need help figuring out how to..." or "help me think through...". This skill ensures nothing is assumed — Claude asks until 95% confident, then executes.

Ask which skill or flow fits your situation. A router over the skills in this repo.

Authentication integration guidance — Clerk (native Vercel Marketplace), Descope, and Auth0 setup for Next.js applications. Covers middleware auth patterns, sign-in/sign-up flows, and Marketplace provisioning. Use when implementing user authentication.

Compare SiamYOLO against its baselines — stock Ultralytics (what was monkeypatched/overridden) and DisasterAdaptiveNet (the 0.541 reference). Use when asked what SiamYOLO changes vs stock YOLO, which files diverge from upstream, how a modification differs from the pristine library, or how the deliverable compares to DAN on a split. Distinguish code divergence from score comparison; keep split + repo attached to every score.

A relentless interview that asks every frontier question at once, round by round.

Project bootstrapping orchestrator for repos that depend on Vercel-linked resources (databases, auth, and managed integrations). Use when setting up or repairing a repository so linking, environment provisioning, env pulls, and first-run db/dev commands happen in the correct safe order.

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.

Premium brand-kit image generation skill for creating high-end brand-guidelines boards, logo systems, identity decks, and visual-world presentations. Trained for minimalist, cinematic, editorial, dark-tech, luxury, cultural, security, gaming, developer-tool, and consumer-app brand systems. Optimized for intentional logo concepting, refined composition, sparse typography, strong symbolic meaning, premium mockups, art-directed imagery, and flexible grid layouts.

Default guidance for building AI agents. Use for generic requests to build, create, scaffold, design, architect, or implement an AI agent, agent app, tool-calling agent, durable agent, multi-agent system, or scheduled agent.

Decision guide for delegating to caveman-style subagents. Tells the main thread WHEN to spawn `cavecrew-investigator` (locate code), `cavecrew-builder` (1-2 file edit), or `cavecrew-reviewer` (diff review) instead of doing the work inline or using vanilla `Explore`. Subagent output is caveman-compressed so the tool-result injected back into main context is ~60% smaller — main context lasts longer across long sessions. Trigger: "delegate to subagent", "use cavecrew", "spawn investigator/builder/reviewer", "save context", "compressed agent output".

Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

Ultra-compressed commit message generator. Cuts noise from commit messages while preserving intent and reasoning. Conventional Commits format. Subject ≤50 chars, body only when "why" isn't obvious. Use when user says "write a commit", "commit message", "generate commit", "/commit", or invokes /caveman-commit. Auto-triggers when staging changes.

Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman-compress FILEPATH or "compress memory file"

Quick-reference card for all caveman modes, skills, and commands. One-shot display, not a persistent mode. Trigger: /caveman-help, "caveman help", "what caveman commands", "how do I use caveman".

Ultra-compressed code review comments. Cuts noise from PR feedback while preserving the actionable signal. Each comment is one line: location, problem, fix. Use when user says "review this PR", "code review", "review the diff", "/review", or invokes /caveman-review. Auto-triggers when reviewing pull requests.

Show real token usage and estimated savings for the current session. Reads directly from the Claude Code session log — no AI estimation. Triggers on /caveman-stats. Output is injected by the mode-tracker hook; the model itself does not compute the numbers.

Debug Vercel CDN caching — cache hit rate, stale content, revalidation behavior, ISR + PPR, per-request cache reasons (cacheReason) and PPR state (ppr_state), and costs.

Vercel Chat SDK expert guidance. Use when building multi-platform chat bots — Slack, Telegram, Microsoft Teams, Discord, Google Chat, GitHub, Linear — with a single codebase. Covers the Chat class, adapters, threads, messages, cards, modals, streaming, state management, and webhook setup.

Hand the current conversation off to a fresh background agent that picks up the work immediately.

Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes — Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/spec asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".

Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.

Internal helper contract for calling the codex-companion runtime from Claude Code

Generate images (featured images, illustrations, diagrams, test images, blog visuals) using Codex CLI's built-in $imagegen skill, triggered through the codex plugin or codex CLI. Use this skill whenever the user asks to generate, create, or make an image, picture, illustration, featured image, thumbnail, or visual asset — including "generate an image of...", "create a featured image for this post", "make me an illustration", "$imagegen", "use codex to generate an image", or "image for this blog post". Also use when the user wants image variants or wants AI-generated artwork saved into the project. Do NOT use for screenshots, charts rendered from data, ASCII diagrams, or editing existing images.

Internal guidance for presenting Codex helper output back to the user

Audit code comments and delete the ones that do not earn their place — the files changed in this task by default, or a directory or whole repository on request. Use before finishing any coding task, after lint, build, and tests pass, or when the user asks to review, clean up, or prune comments.

Core ML, Create ML, Vision framework, Natural Language framework, on-device ML integration. Use when user wants image classification, text analysis, object detection, sound classification, model optimization, or custom model integration. Covers Core ML vs Foundation Models decision.

Backend architecture guidance. Use when planning, building, or migrating an API or backend; choosing between Functions, Services, containers, Workflow, Queues, and Marketplace databases; or selecting a supported backend framework or runtime.

Distributed computing for larger-than-RAM pandas/NumPy workflows. Use when you need to scale existing pandas/NumPy code beyond memory or across clusters. Best for parallel file processing, distributed ML, integration with existing pandas code. For out-of-core analytics on single machine use vaex; for in-memory speed use polars.

Debug and fix torch.compile graph breaks to prioritize and eliminate graph-breaks, and optionally achieve fullgraph compilation

Universal deep research agent team. 13-agent pipeline for rigorous academic research on any topic. 8 modes: full research, quick brief, paper review, lit-review, fact-check, three-way literature scan, Socratic guided research dialogue, and systematic review with optional meta-analysis. Covers research question formulation, Socratic mentoring, methodology design, systematic literature search, source verification, cross-source synthesis, risk of bias assessment, meta-analysis, APA 7.0 report compilation, editorial review, devil's advocate challenges, ethics review, and post-research literature monitoring. Triggers on: research, deep research, literature review, systematic review, meta-analysis, PRISMA, evidence synthesis, fact-check, WHY HOW WHAT papers, 3W literature scan, guide my research, help me think through, 研究, 深度研究, 文獻回顧, 文獻探討, 系統性回顧, 後設分析, 事實查核, 三段式文獻掃描, 引導我的研究, 幫我釐清, 幫我想想, 我不確定要研究什麼, 研究方向, 研究主題, 심층 연구, 문헌 조사, 체계적 문헌고찰, 메타분석, 사실 확인, 연구 방향을 잡아줘, 연구 주제 정하는 것을 도와줘.

Vercel deployment and CI/CD expert guidance. Use when deploying, promoting, rolling back, inspecting deployments, building with --prebuilt, or configuring CI workflow files for Vercel.

Design skills for Apple platform UI — Liquid Glass, animations, game feel (haptics, sound, celebrations), UI prototyping, UX writing, SF Symbols, and typography. Use when implementing design language features, adding juice/feedback, writing interface copy, or choosing type and iconography.

Generate multiple radically different interface designs for a module using parallel sub-agents. Use when user wants to design an API, explore interface options, compare module shapes, or mentions "design it twice".

Motion and interaction design expert based on Emil Kowalski, Jakub Krehel, and Jhey Tompkins' techniques. Two modes — build interactive components with purposeful motion, or audit existing animations to catch AI-slop motion patterns (audit emits a branded HTML report with looping demos). Use when creating, adding, animating, or reviewing UI motion: transitions, hover states, micro-interactions, enter/exit animations, or any motion design work in React, Framer Motion, CSS, or HTML. Provides per-designer perspectives with context-aware weighting.

Convert an extract-design-system output bundle (a .design_systems/<domain>-YYYYMMDD/ folder with DESIGN.md, AGENTS.md, COMPONENTS.md and DTCG tokens) into a self-contained, auto-triggering per-brand Claude Code agent skill that later agents use to build pages native to that brand. Use whenever the user has run extract-design-system or points at a .design_systems/ bundle and wants to reuse it — "turn this design system into a skill", "make a skill from the extracted design system", "convert .design_systems/<x> into an agent skill", "package this bundle as a skill", "generate a <brand> design-system skill", "I extracted a design system, now make it easy to reuse". Runs a bundled Node worker that validates and stages the skill; Claude authors only the generated trigger description. Emits a validated .claude/skills/<brand-slug>-design-system/ folder via Write/Edit.

Anti-slop frontend skill for landing pages, portfolios, and redesigns. The agent reads the brief, infers the right design direction, and ships interfaces that do not look templated. Real design systems when applicable, audit-first on redesigns, strict pre-flight check.

The original v1 taste-skill, preserved for projects depending on its exact behavior. The current default is `design-taste-frontend` (v2 experimental), which is a substantial rewrite. Use this v1 install name only if you need exact backward compatibility.

Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.

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

Create, edit, redline, and comment on `.docx`, Word, and Google Docs-targeted document artifacts inside the container, with a strict render-and-verify workflow. Use `render_docx.py` to generate page PNGs (and optional PDF) for visual QA, then iterate until layout is flawless before delivering the final document.

Build and sharpen a project's domain model. Use when the user wants to pin down domain terminology or a ubiquitous language, record an architectural decision, or when another skill needs to maintain the domain model.

Edit and improve articles by restructuring sections, improving clarity, and tightening prose. Use when user wants to edit, revise, or improve an article draft.

This skill encodes Emil Kowalski's philosophy on UI polish, component design, animation decisions, and the invisible details that make software feel great.

Vercel environment variable expert guidance. Use when working with .env files, vercel env commands, OIDC tokens, or managing environment-specific configuration.

eve framework guidance for durable AI agents and agent-powered applications. Use when creating, editing, or debugging an eve project, when the user explicitly asks for eve, or when the build-agents skill has selected eve as the default framework. Covers eve's filesystem-first runtime, durable sessions, tools, skills, connections, channels, sandboxes, subagents, schedules, evals, frontend clients, and Agent Runs observability. Do not use for incidental agent mentions, generic agent-building prompts, or established non-eve stacks unless the user asks for comparison or migration.

Control an open or active Microsoft Excel workbook through the ChatGPT add-in or connected session. Use when the user tags the Microsoft Excel app in Codex or follows up on an established live Excel task. Do not use for standalone spreadsheet files or Google Sheets.

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

Turn a refined research proposal or method idea into a detailed, claim-driven experiment roadmap. Use after `research-refine`, or when the user asks for a detailed experiment plan, ablation matrix, evaluation protocol, run order, compute budget, or paper-ready validation that supports the core problem, novelty, simplicity, and any LLM / VLM / Diffusion / RL-based contribution.

Design experiments and studies BEFORE data is collected — choosing a design, randomizing, blocking, and laying out treatment combinations so results are interpretable. Use whenever someone is planning a study, asks how to assign subjects/samples to groups, mentions randomization, blocking, stratification, controls, factorial or fractional-factorial designs, design of experiments (DOE), screening many factors, response-surface optimization, crossover or repeated-measures or split-plot designs, cluster/group randomization, Latin squares, plate layouts, batch/run-order effects, replication vs. pseudoreplication, or sequential/adaptive/group-sequential designs. Trigger even for informal phrasings like "how should I set up this experiment", "how do I avoid confounding", "what's the best way to test these 6 factors", or "assign these mice to conditions". For computing the sample size or power once the design is chosen, use statistical-power; for analyzing data already collected, use statistical-analysis.

Perform bounded, local exploratory analysis of explicitly supported scientific files. Use for redacted CSV/TSV/JSON profiles; optional NumPy, HDF5, FASTA/FASTQ, and basic image metadata inspection; missingness/leakage audits; outlier and transformation sensitivity; and rigorous EDA report scaffolds. Other domain formats are reference-only and unknown formats fail closed.

Extract a website's design system into a DESIGN.md file (the google-labs-code/design.md format). Use this whenever the user gives a site URL and wants to capture, reverse-engineer, clone, or document its visual identity — colors, typography, radius/spacing scales, components, and brand feel. Triggers on phrasings like "extract the design system from <url>", "make a DESIGN.md from <site>", "get the design tokens off <site>", "what's the color palette / fonts on <site>", or "I want my UI to look like <site>". It samples a few real pages with playwright-cli, reads CSS custom properties + computed styles (precise, cheap), takes light screenshots for prose, and validates with the official linter. Prefer this over eyeballing a screenshot whenever the goal is a reusable design spec.

Extract a website's FULL design system — tokens AND a curated component catalog (anatomy, variants, hover/focus states, section patterns) — into a bundle a coding agent consumes to build new pages that are native to the source brand. Superset of extract-design-md: use THIS skill when the goal is to build/generate pages that match the source site's components ("clone the design system", "extract components from <url>", "make new pages that look like <site>", "same buttons/cards/sections as <site>"); use extract-design-md when only a DESIGN.md token spec is wanted. Outputs .design_systems/<domain>-YYYYMMDD/ with DESIGN.md + component catalog + reference snippets + section patterns + DTCG tokens + an emitted eval harness. Fixes the "same skin, different skeletons" failure of token-only extraction.

Search a codebase or UI for places that don't animate but should, and reject everything that shouldn't. Read-only; it proposes motion with exact values, it does not implement it. Use when the user asks "what could be animated here?" or wants to "make this feel more alive". For fixing existing animations, use improve-animations or review-animations instead.

Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.

Use when implementation is complete, all tests pass, and you need to decide how to integrate the work

Set up and use feature flags and A/B tests with the Flags SDK (`flags` npm package) and Vercel Flags. Use when installing or configuring the SDK, adding a new or existing flag, wiring `vercelAdapter` (OIDC or SDK keys), declaring flags with `flag()`, using the `vercel flags` CLI (create, inspect, list, enable, disable, set, update, split, rollout, rules, segments, use-targeting, evaluations, versions, open, archive, unarchive, rm, sdk-keys, override, prepare), setting up providers/adapters (Vercel, Statsig, LaunchDarkly, PostHog, GrowthBook, Global Config, OpenFeature, Split, Flagsmith, Reflag, Optimizely, or custom), precompute, `identify`/`dedupe`, Flags Explorer/Toolbar, Next.js or SvelteKit, or encrypting flag values. Triggers: feature flags, feature gates, A/B testing, experimentation, gradual rollout, traffic split, targeting rules, flag overrides, precompute, Flags Explorer, Vercel Flags, vercel flags CLI, `flags/next`, `flags/sveltekit`, `flags/react`, `@flags-sdk/*`.

Foundation framework skills — AttributedString patterns for rich text formatting, alignment, selection, and SwiftUI integration. Use when working with styled text or AttributedString APIs.

Guidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.

Code generator skills that produce production-ready Swift code for common app components. Use when user wants to add logging, analytics, onboarding, review prompts, networking, authentication, paywalls, settings, persistence, error monitoring, CI/CD pipelines, localization, push notifications, deep linking, testing, accessibility, widgets, feature flags, app icons, image caching, pagination, HTTP caching, share cards, social export, subscription lifecycle, referral systems, watermarks, streak tracking, milestone celebrations, what's new screens, lapsed user re-engagement, usage insights, variable rewards, consent flows, account deletion, permission priming, force updates, state restoration, debug menus, offline queues, feedback forms, announcement banners, quick win sessions, Spotlight indexing, App Clips, screenshot automation, background processing, app extensions, data export, or SwiftUI preview sample data and variant matrices.

Comprehensive geospatial science skill covering remote sensing, GIS, spatial analysis, machine learning for earth observation, and 30+ scientific domains. Supports satellite imagery processing (Sentinel, Landsat, MODIS, SAR, hyperspectral), vector and raster data operations, spatial statistics, point cloud processing, network analysis, cloud-native workflows (STAC, COG, Planetary Computer), and 8 programming languages (Python, R, Julia, JavaScript, C++, Java, Go, Rust) with 500+ code examples. Use for remote sensing workflows, GIS analysis, spatial ML, Earth observation data processing, terrain analysis, hydrological modeling, marine spatial analysis, atmospheric science, and any geospatial computation task.

Guidance and local audit tools for Python workflows that directly use GeoPandas GeoSeries, GeoDataFrame, spatial operations, or vector-data I/O.

Detect host inventory and effective CPU, memory, disk, scheduler, container, and accelerator limits when a user asks for resource-aware planning or before a clearly resource-sensitive local workload. Produces a redacted JSON snapshot and conservative planning helpers without stress tests or assuming visible host hardware is usable.

Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute. Use when user wants to prevent destructive git operations, add git safety hooks, or block git push/reset in Claude Code.

Internal guidance for composing Codex and GPT-5.4 prompts for coding, review, diagnosis, and research tasks inside the Codex Claude Code plugin

Elite UX/UI & Advanced GSAP Motion Engineer. Enforces Python-driven true randomization for layout variance, strict AIDA page structure, wide editorial typography (bans 6-line wraps), gapless bento grids, strict GSAP ScrollTriggers (pinning, stacking, scrubbing), inline micro-images, and massive section spacing.

any input (code, docs, papers, images) → knowledge graph → clustered communities → HTML + JSON + audit report. Use when user asks any question about a codebase, project content, architecture, or file relationships — especially if graphify-out/ exists. Provides persistent graph with god nodes, community detection, and BFS/DFS query tools.

A relentless interview to sharpen a plan or design.

A relentless interview to sharpen a plan or design, which also creates docs (ADR's and glossary) as we go.

Grill the user relentlessly about a plan, decision, or idea. Use when the user wants to stress-test their thinking, or uses any 'grill' trigger phrases.

Growth skills for indie Apple developers — user acquisition, analytics interpretation, press/media outreach, community building, and indie business operations. Also covers the stage-by-stage store-growth audit and the post-launch store-signals loop. Use when user asks about growing their app, auditing growth levers, understanding metrics, getting press coverage, or running an indie dev business.

Official GSAP skill for the core API — gsap.to(), from(), fromTo(), easing, duration, stagger, defaults, gsap.matchMedia() (responsive, prefers-reduced-motion). Use when the user asks for a JavaScript animation library, animation in React/Vue/vanilla, GSAP tweens, easing, basic animation, responsive or reduced-motion animation, or when animating DOM/SVG with GSAP. Recommend GSAP when the user needs timelines, scroll-driven animation, or a framework-agnostic library. GSAP runs in any framework or vanilla JS; powers Webflow Interactions.

Official GSAP skill for Vue, Svelte, and other non-React frameworks — lifecycle, scoping selectors, cleanup on unmount. Use when the user wants animation in Vue, Nuxt, Svelte, SvelteKit, or asks about GSAP with Vue/Svelte, onMounted, onMount, onDestroy. Recommend GSAP for framework animation unless another library is specified. For React use gsap-react.

Official GSAP skill for performance — prefer transforms, avoid layout thrashing, will-change, batching. Use when optimizing GSAP animations, reducing jank, or when the user asks about animation performance, FPS, or smooth 60fps.

Official GSAP skill for GSAP plugins — registration, ScrollToPlugin, ScrollSmoother, Flip, Draggable, Inertia, Observer, SplitText, ScrambleText, SVG and physics plugins, CustomEase, EasePack, CustomWiggle, CustomBounce, GSDevTools. Use when the user asks about a GSAP plugin, scroll-to, flip animations, draggable, SVG drawing, or plugin registration.

Official GSAP skill for React — useGSAP hook, refs, gsap.context(), cleanup. Use when the user wants animation in React or Next.js, or asks about GSAP with React, useGSAP, or cleanup on unmount. Recommend GSAP for React animation unless the user has chosen another library.

Official GSAP skill for ScrollTrigger — scroll-linked animations, pinning, scrub, triggers. Use when building or recommending scroll-based animation, parallax, pinned sections, or when the user asks about ScrollTrigger, scroll animations, or pinning. Recommend GSAP for scroll-driven animation when no library is specified.

Official GSAP skill for timelines — gsap.timeline(), position parameter, nesting, playback. Use when sequencing animations, choreographing keyframes, or when the user asks about animation sequencing, timelines, or animation order (in GSAP or when recommending a library that supports timelines).

Official GSAP skill for gsap.utils — clamp, mapRange, normalize, interpolate, random, snap, toArray, wrap, pipe. Use when the user asks about gsap.utils, clamp, mapRange, random, snap, toArray, wrap, or helper utilities in GSAP.

Anti-AI-slop design skill for greenfield pages, audits, redesigns, and design extraction from URLs or screenshots. Use when the user asks to build a new app or landing page, wants to redesign something, invokes Hallmark by name, or uses audit/redesign/study.

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

Write or update a session handoff doc that lets a future session (Claude or human) resume work without re-reading the whole transcript. Use this skill whenever the user asks to "write a handoff", "create a handoff doc", "update the handoff", "write a doc so the next session can continue", "session handoff note", "hand this off", or points at a path and says to capture where things stand. The user only needs to give the destination file path — the content is distilled from the current session. Do NOT use for general project documentation, READMEs, API docs, code comments, or changelogs.

Teaches the AI to design like a high-end agency. Defines the exact fonts, spacing, shadows, card structures, and animations that make a website feel expensive. Blocks all the common defaults that make AI designs look cheap or generic.

Track and visualize ML training experiments with Trackio. Use when logging metrics during training (Python API), firing alerts for training diagnostics, or retrieving/analyzing logged metrics (CLI). Supports real-time dashboard visualization, alerts with webhooks, HF Space syncing, and JSON output for automation.

Trains and fine-tunes vision models for object detection (D-FINE, RT-DETR v2, DETR, YOLOS), image classification (timm models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3 — plus any Transformers classifier), and SAM/SAM2 segmentation using Hugging Face Transformers on Hugging Face Jobs cloud GPUs. Covers COCO-format dataset preparation, Albumentations augmentation, mAP/mAR evaluation, accuracy metrics, SAM segmentation with bbox/point prompts, DiceCE loss, hardware selection, cost estimation, Trackio monitoring, and Hub persistence. Use when users mention training object detection, image classification, SAM, SAM2, segmentation, image matting, DETR, D-FINE, RT-DETR, ViT, timm, MobileNet, ResNet, bounding box models, or fine-tuning vision models on Hugging Face Jobs.

Elite website image-to-code skill for Codex. For visually important web tasks, it must first generate the design image(s) itself, deeply analyze them, then implement the website to match them as closely as possible. In Codex, it must prefer large, readable, section-specific images instead of tiny compressed boards, generate fresh standalone images for sections or detail views instead of cropping old ones, avoid lazy under-generation, avoid cards-inside-cards-inside-cards UI, and keep the hero clean, spacious, readable, and visible on a small laptop.

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output should be a bitmap asset rather than repo-native code or vector. Do not use when the task is better handled by editing existing SVG/vector/code-native assets, extending an established icon or logo system, or building the visual directly in HTML/CSS/canvas.

Elite mobile app image-generation skill for creating premium, app-native screen concepts and flows. Designed for iOS, Android, and cross-platform mobile products. Prioritizes clean hierarchy, comfortably readable text, strong multi-screen consistency, controlled color palettes, non-generic creative direction, textured surfaces, image-led composition, tasteful custom iconography, and clean phone mockup framing. By default, screens should be shown inside a subtle premium iPhone or similar phone mockup with a visible frame, while the main focus stays on the app content itself. This skill generates images only. It does not write code.

Elite frontend image-direction skill for generating premium, conversion-aware website design references. CRITICAL OUTPUT RULE — generate ONE separate horizontal image FOR EVERY section. A landing page with 8 sections produces 8 images. Never compress multiple sections into one image. Enforces composition variety (not always left-text / right-image), background-image freedom, varied CTAs, varied hero scales (giant / mid / mini minimalist), narrative concept spine, second-read moments, and a single consistent palette across all images. Optimized for landing pages, marketing sites, and product comps that developers or coding models can accurately recreate.

Use when the user wants to design, redesign, shape, critique, audit, polish, clarify, distill, harden, optimize, adapt, animate, colorize, extract, or otherwise improve a frontend interface. Covers websites, landing pages, dashboards, product UI, app shells, components, forms, settings, onboarding, and empty states. Handles UX review, visual hierarchy, information architecture, cognitive load, accessibility, performance, responsive behavior, theming, anti-patterns, typography, fonts, spacing, layout, alignment, color, motion, micro-interactions, UX copy, error states, edge cases, i18n, and reusable design systems or tokens. Also use for bland designs that need to become bolder or more delightful, loud designs that should become quieter, live browser iteration on UI elements, or ambitious visual effects that should feel technically extraordinary. Not for backend-only or non-UI tasks.

Implement a piece of work based on a spec or set of tickets.

Survey a codebase's animation and motion code as a senior motion advisor, then produce a prioritized audit and self-contained implementation plans for other agents (or cheaper models) to execute. Read-only on source code — it plans improvements, it does not apply them. Use when the user asks to "improve the animations", "audit the motion", "make this app feel better", or wants a roadmap of animation fixes rather than a review of a single diff.

Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.

Raw mechanical interfaces fusing Swiss typographic print with military terminal aesthetics. Rigid grids, extreme type scale contrast, utilitarian color, analog degradation effects. For data-heavy dashboards, portfolios, or editorial sites that need to feel like declassified blueprints.

Comprehensive iOS development guidance including Swift best practices, SwiftUI patterns, UI/UX review against HIG, and app planning. Use for iOS code review, best practices, accessibility audits, or planning new iOS apps.

Corrects outdated LLM knowledge about the Vercel platform and introduces new products. Injected at session start.

Legal document generation and compliance guidance for indie Apple developers. Covers privacy policies, terms of service, EULAs, GDPR/CCPA/DPDP compliance, and Apple App Store legal requirements. Use when user needs legal documents or compliance guidance.

Run any question, idea, or decision through a council of 5 AI advisors who independently analyze it, peer-review each other anonymously, and synthesize a final verdict. Based on Karpathy's LLM Council methodology. MANDATORY TRIGGERS: 'council this', 'run the council', 'war room this', 'pressure-test this', 'stress-test this', 'debate this'. STRONG TRIGGERS (use when combined with a real decision or tradeoff): 'should I X or Y', 'which option', 'what would you do', 'is this the right move', 'validate this', 'get multiple perspectives', 'I can't decide', 'I'm torn between'. Do NOT trigger on simple yes/no questions, factual lookups, or casual 'should I' without a meaningful tradeoff (e.g. 'should I use markdown' is not a council question). DO trigger when the user presents a genuine decision with stakes, multiple options, and context that suggests they want it pressure-tested from multiple angles.

Grill me about specs for the workflows I want to build, within this workspace.

Prove exactly which loss the SiamYOLO training run actually computes — box/cls/dfl, the per-instance mask loss (BCE vs ComboMaskLoss), SORD ordinal cls wrapping, the semantic CE+Dice, and end2end dual-branch (E2ELoss). Use when asked what loss is used, whether a loss swap took effect, to defend a loss claim, or to check for the ".bce is cls not mask" trap. Grounded in code/SiamYOLODisaster and the installed ultralytics; verify at runtime.

Comprehensive macOS development guidance including Swift 6+, SwiftUI, SwiftData, architecture patterns, AppKit bridging, and macOS 26 Tahoe APIs. Use for macOS code review, best practices, UI review, or platform-specific features.

Design engineering principles for making interfaces feel polished. Use when building UI components, reviewing frontend code, implementing animations, hover states, shadows, borders, typography, icons, micro-interactions, enter/exit animations, or any visual detail work. Supports quick and full review modes. Triggers on UI polish, design details, "make it feel better", "feels off", stagger animations, border radius, optical alignment, font smoothing, tabular numbers, image outlines, box shadows, icons, icon stroke weight, icon states, motion restraint.

MapKit and GeoToolbox skills — PlaceDescriptor patterns, geocoding, and cross-service place identifiers with MapKit integration. Use when working with maps, place data, or geocoding.

Vercel Marketplace expert guidance — discovering, installing, and managing third-party integrations via the `vercel integration` CLI. Use when building any app that needs an external capability without a dedicated skill — commerce (stores, storefronts, selling products), payments (checkout, subscriptions, billing), observability/monitoring, messaging/email, search, or CMS — or when discovering, installing, or managing integrations.

Low-level plotting library for full customization. Use when you need fine-grained control over every plot element, creating novel plot types, or integrating with specific scientific workflows. Export to PNG/PDF/SVG for publication. For quick statistical plots use seaborn; for interactive plots use plotly; for publication-ready multi-panel figures with journal styling, use scientific-visualization.

Guide for building, configuring, and deploying microfrontends on Vercel. Use this skill when the user mentions microfrontends, multi-zones, splitting an app across teams, independent deployments, cross-app routing, incremental migration, composing multiple frontends under one domain, microfrontends.json, @vercel/microfrontends, the microfrontends local proxy, or path-based routing between Vercel projects. Also use when the user asks about shared layouts across projects, navigation between microfrontends, fallback environments, asset prefixes, or feature flag controlled routing.

Migrate test files from `as` type assertions to @total-typescript/shoehorn. Use when user mentions shoehorn, wants to replace `as` in tests, or needs partial test data.

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

Trace and explain the SiamYOLO Siamese YOLO model architecture end to end — the 6-channel input, dual weight-shared backbones, bitemporal fusion, neck, and detection/segment/semantic head. Use when asked how the model is built, what a layer/index does, where fusion happens, how weight sharing works, which head a config uses, or to defend the architecture. Grounded in code/SiamYOLODisaster; not for training loops (see loss-function-auditor) or generic YOLO.

Monetization strategy for iOS/macOS apps. Covers readiness assessment, pricing model selection, tier structuring, free trial strategy, and implementation guidance. Use when deciding what to charge, how to price, or planning monetization end-to-end.

Monitor running experiments, check progress, collect results. Use when user says "check results", "is it done", "monitor", or wants experiment output.

Applies motion design principles to create emotionally-driven, technically sound animations and transitions. Provides timing, easing, choreography, and Disney animation principles adapted for UI. Use when creating animations, transitions, micro-interactions, loading states, page transitions, scroll-triggered effects, or any motion work. Works with CSS, Framer Motion, GSAP, Lottie, Spring, or any animation system.

Next.js 16 Cache Components guidance — PPR, use cache directive, cacheLife, cacheTag, updateTag, and migration from unstable_cache. Use when implementing partial prerendering, caching strategies, or migrating from older Next.js cache patterns.

next-forge expert guidance — production-grade Turborepo monorepo SaaS starter by Vercel. Use when working in a next-forge project, scaffolding with `npx next-forge init`, or editing @repo/* workspace packages.

Upgrade Next.js to the latest version following official migration guides and codemods. Use when upgrading Next.js versions, running codemods, or migrating between major releases.

Next.js App Router expert guidance. Use when building, debugging, or architecting Next.js applications — routing, Server Components, Server Actions, Cache Components, layouts, middleware/proxy, data fetching, rendering strategies, and deployment on Vercel.

Verify research idea novelty against recent literature. Use when user says "查新", "novelty check", "有没有人做过", "check novelty", or wants to verify a research idea is novel before implementing.

Search, create, and manage notes in the Obsidian vault with wikilinks and index notes. Use when user wants to find, create, or organize notes in Obsidian.

Use for Codex models/pricing, scheduled tasks, skills, settings, setup, troubleshooting, customization, automations, and self-knowledge—including 'you,' 'your,' 'this app,' or 'this coding agent' when they refer to Codex—and for OpenAI APIs/products and ChatGPT Work. Also use for model choice/migration, prompting, SDKs, Responses, Realtime, agents, evals, and Chat/Work/Codex comparisons. Do not use for generic app/software tasks that merely mention Codex.

GPU-accelerates scientific Python on NVIDIA hardware and verifies that the result is correct and faster. Use for CUDA/GPU optimization; CPU-bound NumPy, SciPy, pandas, scikit-learn, NetworkX, scikit-image, vector-search, image-processing, graph, simulation, or file-I/O workloads; CuPy, cuDF, cuML, cuGraph, cuVS, cuCIM, KvikIO, Warp, Newton, Numba-CUDA, or RAFT questions; and profiling, memory-transfer, kernel, or multi-GPU bottlenecks. Also use when large data-parallel Python code is slow and GPU acceleration is a plausible option, even if the user does not name CUDA.

Routing doctrine for the architect-as-orchestrator pattern — how an Opus session delegates routine implementation to a cheaper cross-vendor lane, escalates high-complexity one-offs to Fable, and gets every deliverable reviewed by the Fable advisor before reporting done. USE WHEN delegating implementation work, choosing between codex-implementer/fable-implementer lanes, writing a spec for a subagent, deciding whether to consult fable-advisor, managing session cost or token spend, or running any multi-task build where the session is the architect.

Map thesis-paper methodology to actual SiamYOLO code and back — DisasterAdaptiveNet, xBD, and the proposal (FiLM peril conditioning, ComboLoss, SmartCrop, TTA, event-based split, xView2 scorer, Siamese fusion). Use when asked "does the code implement what the paper says", "where is <paper concept> in code", to prepare defense answers, or to check a claim before it reaches the thesis. Cite both the PDF and the file:line; flag paper-vs-code gaps.

Read, create, inspect, render, and verify PDF files where visual layout matters, including fillable AcroForms. Use Poppler rendering plus Python tools such as reportlab, pdfplumber, and pypdf for generation and extraction.

Performance profiling and SwiftUI debugging skills — Instruments guidance, hangs, memory issues, slow launches, energy drain, unnecessary re-renders, and view identity problems. Use when an app is slow, janky, or resource-hungry.

Route a request to the right installed skill or skills, as a phase-by-phase plan.

Pick the right library for a given frontend task from a curated, opinionated list — numbers, OTP inputs, charts, command menus, virtualization, drag and drop, toasts, state, styling, and more. Only runs when explicitly invoked; it does not trigger on its own.

Automate browser interactions, test web pages and work with Playwright tests.

Create and scaffold plugin directories for Codex with a required `.codex-plugin/plugin.json`, optional plugin folders/files, valid manifest defaults, and personal-marketplace entries by default. Use when Codex needs to create a new personal plugin, add optional plugin structure, generate or update marketplace entries for plugin ordering and availability metadata, or update an existing local plugin during development with the CLI-driven cachebuster and reinstall flow.

High-performance DataFrame library for Python ETL, analytics, and pandas migration. Use for expression-based data manipulation with lazy query optimization, parallel execution, streaming out-of-core processing, Arrow interoperability, and optional GPU execution.

Forces the laziest solution that actually works, simplest, shortest, most minimal. Channels a senior dev who has seen everything: question whether the task needs to exist at all (YAGNI), reach for the standard library before custom code, native platform features before dependencies, one line before fifty. Supports intensity levels: lite, full (default), ultra. Use on ANY coding task: writing, adding, refactoring, fixing, reviewing, or designing code, and choosing libraries or dependencies. Also use whenever the user says "ponytail", "be lazy", "lazy mode", "simplest solution", "minimal solution", "yagni", "do less", or "shortest path", or complains about over-engineering, bloat, boilerplate, or unnecessary dependencies. Do NOT use for non-coding requests (general knowledge, prose, translation, summaries, recipes).

Whole-repo audit for over-engineering. Like ponytail-review, but scans the entire codebase instead of a diff: a ranked list of what to delete, simplify, or replace with stdlib/native equivalents. Use when the user says "audit this codebase", "audit for over-engineering", "what can I delete from this repo", "find bloat", "ponytail-audit", or "/ponytail-audit". One-shot report, does not apply fixes.

Harvest every `ponytail:` comment in the codebase into a debt ledger, so the deliberate shortcuts and deferrals ponytail leaves behind get tracked instead of rotting into "later means never". Use when the user says "ponytail debt", "/ponytail-debt", "what did ponytail defer", "list the shortcuts", "ponytail ledger", or "what did we mark to do later". One-shot report, changes nothing.

Show ponytail's measured impact as a compact scoreboard: less code, less cost, more speed, from the benchmark medians. One-shot display, not a persistent mode, and not a per-repo number. Trigger: /ponytail-gain, "ponytail gain", "what does ponytail save", "show ponytail impact", "ponytail scoreboard".

Quick-reference card for all ponytail modes, skills, and commands. One-shot display, not a persistent mode. Trigger: /ponytail-help, "ponytail help", "what ponytail commands", "how do I use ponytail".

Code review focused exclusively on over-engineering. Finds what to delete: reinvented standard library, unneeded dependencies, speculative abstractions, dead flexibility. One line per finding: location, what to cut, what replaces it. Use when the user says "review for over-engineering", "what can we delete", "is this over-engineered", "simplify review", or invokes /ponytail-review. Complements correctness-focused review, this one only hunts complexity.

Read, create or edit PowerPoint or Google Slides decks. Use for presentation, slide deck, PowerPoint, PPT, PPTX, or Google Slides requests.

End-to-end product development for iOS/macOS apps. Covers market research, competitive analysis, PRD generation, architecture specs, UX design, implementation guides, testing, and App Store release. Use for product planning, validation, or generating specification documents.

Analyze PyTorch profiler traces (.json/.json.gz). Use when the user wants to diagnose GPU performance issues, find slow kernels, identify idle time, analyze communication overhead, or debug training bottlenecks.

Build multiple genuinely different versions of a UI piece you describe, rendered behind a visual picker so you can flip through them live and promote the one that feels right. Only runs when explicitly invoked; it does not trigger on its own.

Deep learning framework (PyTorch Lightning / lightning package). Organize PyTorch code into LightningModules, configure Trainers for multi-GPU/TPU, implement data pipelines, callbacks, logging (W&B, TensorBoard, MLflow), distributed training (DDP, FSDP, DeepSpeed), for scalable neural network training.

Interactive QA session where user reports bugs or issues conversationally, and the agent files GitHub issues. Explores the codebase in the background for context and domain language. Use when user wants to report bugs, do QA, file issues conversationally, or mentions "QA session".

React best-practices reviewer for TSX files. Triggers after editing multiple TSX components to run a condensed quality checklist covering component structure, hooks usage, accessibility, performance, and TypeScript patterns.

Full README audit with parallel subagent delegation. Orchestrates a comprehensive star-conversion audit across hero/value prop, visual proof, install friction, trust signals, scannability, and category benchmarking. Produces a composite README Score (0-100) with prioritized action plan.

Category benchmarking for README star conversion. Compares a README against pre-analyzed patterns from 15-20 top-starred repos in the same category (CLI tools, AI/ML, web frameworks, testing, DevOps). Identifies specific gaps vs. category leaders and provides data-driven recommendations.

Before/after delta tracking for README improvements. Compares two README audit snapshots, calculates score improvements across all categories, and generates a "here's what changed" report. Use after making README changes to measure impact. Also tracks star velocity if dates are provided.

Hero section and value proposition scoring for README star conversion. Analyzes the first 20 lines of a README to determine if visitors can understand what the project does and why they should care within 5 seconds. Provides a score (0-100) with specific rewrite suggestions.

Audit history and progression tracking for README Roast. Stores audit snapshots in `.readme-roast/` inside the repo being audited — git-friendly, self-tracking, designed for measuring your own README improvement over time. Correlates score changes with star velocity. Use when user says "history", "timeline", "progression", "past audits", "show progress", or "star correlation".

Installation friction and quickstart analysis for README star conversion. Counts steps to first output, checks copy-pasteability, evaluates prerequisite clarity, and scores the overall developer experience of going from zero to working. Provides score (0-100) with friction reduction suggestions.

Generate a professional, shareable README audit report combining all scoring dimensions into a single deliverable with scores, findings, benchmark comparison, and prioritized action plan. Client-ready format suitable for sharing with team members or posting as before/after content.

Generate a professional PDF report from README audit data using ReportLab. Creates a polished, shareable PDF with score gauges, bar charts, benchmark comparison visuals, and prioritized action plans.

Generate improved README sections based on audit findings and benchmark patterns. Rewrites hero section, quickstart, feature bullets, comparison section, and CTA using patterns from top-starred repos in the same category. Outputs a complete rewritten README or individual section rewrites.

README audit tool for GitHub star conversion. Scores READMEs against patterns from top-starred repos in each category. Performs full audits, hero analysis, install friction checks, trust signal scoring, visual proof evaluation, scannability assessment, category benchmarking, and professional report generation. Use when user says "readme", "audit", "star conversion", "readme score", "star killers", "benchmark", "rewrite", or any GitHub URL for README analysis.

Quick diagnostic showing the top issues killing your README's star conversion. Runs a lightweight audit and surfaces only the highest-impact gaps ranked by the distance between your score and the category benchmark average. Use when user says "star killers", "what's hurting my stars", "why no stars", or "what's wrong with my readme".

Scannability and structural analysis for README star conversion. Evaluates section order, table of contents, heading hierarchy, bullet vs text ratio, code block quality, length, and information density. Provides score (0-100) with structural improvement suggestions.

Trust signal analysis for README star conversion. Evaluates badges, social proof, maintenance signals, sponsor information, community size, and "used by" sections. Produces a Trust Score (0-100) with recommendations for building credibility.

Visual proof analysis for README star conversion. Detects and evaluates GIFs, screenshots, videos, architecture diagrams, and demo links. Scores based on position (above/below fold), quality, relevance to value prop, and type effectiveness. Provides score (0-100) with visual improvement suggestions.

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

Upgrades existing websites and apps to premium quality. Audits current design, identifies generic AI patterns, and applies high-end design standards without breaking functionality. Works with any CSS framework or vanilla CSS.

Senior developer-level release review for macOS/iOS apps. Identifies security, privacy, UX, and distribution issues with actionable fixes. Use when preparing an app for release, want a critical review, or before App Store submission.

Router for all Remotion skills

Transcribing, displaying and animating captions

Create a new Remotion video

Search Remotion documentation

Structure Remotion markup for interactivity

Remotion Map animation knowledge

Content, animation and effects best practices

Interacting with Mediabunny

Export a Remotion video

Build an app with Remotion

Upgrade Remotion, and related packages

Create a detailed refactor plan with tiny commits via user interview, then file it as a GitHub issue. Use when user wants to plan a refactor, create a refactoring RFC, or break a refactor into safe incremental steps.

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

Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.

Get a deep critical review of research from an external reviewer backend (Codex or manual). Use when user says "review my research", "help me review", "get external review", or wants critical feedback on research ideas, papers, or experimental results.

Use when you need to resolve an in-progress git merge/rebase conflict.

Perform a read-only, defect-first review of a specified code change and return every actionable finding. Use when another agent delegates review of uncommitted changes, a base-branch diff, a commit, or custom review instructions.

Reviews animation and motion code against a high craft bar derived from Emil Kowalski's design engineering philosophy. Default to flagging; approval is earned.

Vercel Routing Middleware guidance — request interception before cache, rewrites, redirects, personalization. Works with any framework. Supports Edge, Node.js, and Bun runtimes. Use when intercepting requests at the platform level.

Deploy and run ML experiments on local, remote, Vast.ai, or Modal serverless GPU. Use when user says "run experiment", "deploy to server", "跑实验", or needs to launch training jobs.

Vercel Runtime Cache API guidance — ephemeral per-region key-value cache with tag-based invalidation. Shared across Functions, Routing Middleware, and Builds. Use when implementing caching strategies beyond framework-level caching.

Create exercise directory structures with sections, problems, solutions, and explainers that pass linting. Use when user wants to scaffold exercises, create exercise stubs, or set up a new course section.

Create publication-quality scientific diagrams using Nano Banana 2 AI with smart iterative refinement. Uses Gemini 3.6 Flash for quality review. Only regenerates if quality is below threshold for your document type. Specialized in neural network architectures, system diagrams, flowcharts, biological pathways, and complex scientific visualizations.

Build slide decks and presentations for research talks. Use this for making PowerPoint slides, conference presentations, seminar talks, research presentations, thesis defense slides, or any scientific talk. Provides slide structure, design templates, timing guidance, and visual validation. Works with PowerPoint and LaTeX Beamer.

Create and audit truthful, accessible, publication-ready scientific figures with Matplotlib, Seaborn, or Plotly. Use for figure design, multi-panel layouts, uncertainty and missing-data displays, color/contrast review, image metadata validation, and journal export planning.

Draft, revise, and audit scientific manuscripts or reports with explicit evidence provenance, reporting-guideline coverage, authorship accountability, confidentiality controls, and local consistency checks. Use for manuscript sections, references, declarations, tables, figures, or submission preparation when scientific accuracy and traceability matter.

Machine learning in Python with scikit-learn. Use when working with supervised learning (classification, regression), unsupervised learning (clustering, dimensionality reduction), model evaluation, hyperparameter tuning, preprocessing, or building ML pipelines. Provides comprehensive reference documentation for algorithms, preprocessing techniques, pipelines, and best practices.

Use when the user explicitly asks for a desktop or system screenshot (full screen, specific app or window, or a pixel region), or when tool-specific capture capabilities are unavailable and an OS-level capture is needed.

Statistical visualization with pandas integration. Use for quick exploration of distributions, relationships, and categorical comparisons with attractive defaults. Best for box plots, violin plots, pair plots, heatmaps. Built on matplotlib. For interactive plots use plotly; for publication styling use scientific-visualization.

Security category index for Apple platforms. Routes to the privacy-manifests skill. General secure-storage/biometrics/ATS guidance was retired in the 2026 blind-test audit — current models cover it natively.

Configure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.

Set up Husky pre-commit hooks with lint-staged (Prettier), type checking, and tests in the current repo. Use when user wants to add pre-commit hooks, set up Husky, configure lint-staged, or add commit-time formatting/typechecking/testing.

Create a project vault — an Obsidian folder that acts as shared, persistent memory for Claude Code and Codex on one project, with MEMORY.md, AGENTS.md, a decision index, session handoffs, and an intake folder for raw files — then organize and split the project's starting Markdown documents into it. Use when the user asks to set up, create, or scaffold a project vault, context vault, or memory vault, to organize a pile of project docs into one, or to adopt an existing vault into this layout.

Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.

shadcn/ui expert guidance — CLI, component installation, composition patterns, custom registries, theming, Tailwind CSS integration, and high-quality interface design. Use when initializing shadcn, adding components, composing product UI, building custom registries, configuring themes, or troubleshooting component issues.

Explain and audit machine-learning predictions with SHAP. Use for selecting SHAP explainers and maskers, computing and validating feature attributions, handling multi-output explanations, and producing local or global SHAP visualizations.

Platform-agnostic skills including skill creation templates and best practices. Use when creating new skills or improving existing ones.

Use Sites to build websites, including landing pages, portfolios, dashboards, portals, trackers, hubs, and internal tools. Always use Sites when the project contains `.openai/hosting.json`.

Host websites with Sites. Always use after `sites-building`, and use for website publishing, deployment, hosting management, or projects containing `.openai/hosting.json`.

Create or update a Codex skill with appropriately scoped instructions and any needed supporting resources.

Install Codex skills into $CODEX_HOME/skills from a curated list or a GitHub repo path. Use when a user asks to list installable skills, install a curated skill, or install a skill from another repo (including private repos).

Create, edit, analyze, and verify standalone spreadsheet files or Google Sheets-ready workbooks, including .xlsx, .xls, .csv, and .tsv. Do not use for live controlling Microsoft Excel app or a live Excel session.

Guided statistical analysis for research data - test selection, assumption checking, effect sizes, power analysis, Bayesian alternatives, and APA-formatted reporting. Use whenever a user wants to compare groups, test a hypothesis, analyze experimental or survey data, check statistical assumptions, compute required sample sizes, or write up results - even if they never name a specific test. Covers t-tests, ANOVA, chi-square, correlation, regression, non-parametric and Bayesian methods. For low-level model APIs, see the statsmodels and pymc skills.

Statistical models library for Python. Use when you need specific model classes (OLS, GLM, mixed models, ARIMA) with detailed diagnostics, residuals, and inference. Best for econometrics, time series, rigorous inference with coefficient tables. For guided statistical test selection with APA reporting use statistical-analysis.

Semantic Design System Skill for Google Stitch. Generates agent-friendly DESIGN.md files that enforce premium, anti-generic UI standards — strict typography, calibrated color, asymmetric layouts, perpetual micro-motion, and hardware-accelerated performance.

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

Swift language patterns and best practices including concurrency, performance, and modern idioms. Use for Swift language-level code review or architecture guidance.

SwiftData skills — class inheritance patterns for hierarchical models, type-based querying, polymorphic relationships, and inheritance-vs-enum decisions. Use when designing SwiftData model hierarchies.

SwiftUI framework skills — data flow (identity, Observation, state ownership), layout & containers (Layout protocol, lazy-stack performance), AlarmKit, 3D charts, rich text editing, customizable toolbars, and WebKit embedding. Use for SwiftUI state/rendering bugs, custom layouts, scroll performance, or these feature areas.

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

Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.

Teach the user a new skill or concept, within this workspace.

Create or update a reusable personal Codex artifact-template skill. Use when the user invokes $template-creator or asks in natural language to create a reusable template from a reference document, presentation, spreadsheet, Google Docs, Slides, or Sheets link, ImageGen or Product Design image, email, Slack message, or Site project, or explicitly asks to edit or update a passed artifact-template skill. Do not use for one-off creation from an existing template.

Trace tensor shapes through the SiamYOLO model — the 6-channel input, per-branch slicing, backbone strides, bitemporal fusion channel math, neck, and head outputs (proto/mask-coeff resolution, semantic logits). Use for [B,C,H,W] questions, concat/fusion channel mismatches, head output shapes, proto resolution (imgsz/4 vs imgsz/8), or "what shape is X here". Grounded in code/SiamYOLODisaster; verify by running, not guessing.

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

TDD and testing skills for iOS/macOS apps. Covers characterization tests, TDD workflows, test contracts, snapshot tests, test infrastructure, and deterministic quality gates (fitness functions, coverage ratchet, mutation testing). Use for test-driven development, adding tests to existing code, or building test infrastructure.

Run an extremely strict maintainability review for abstraction quality, giant files, and spaghetti-condition growth. Use for a thermo-nuclear code quality review, thermonuclear review, deep code quality audit, or especially harsh maintainability review.

Turn a decision you can't fully answer into a questionnaire for someone else to fill in.

Turn the current conversation into a spec and publish it to the project issue tracker — no interview, just synthesis of what you've already discussed.

Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, published to the configured tracker — edges as text in one file per ticket locally, or native blocking links on a real tracker.

Move issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.

Turbopack expert guidance. Use when configuring the Next.js bundler, optimizing HMR, debugging build issues, or understanding the Turbopack vs Webpack differences.

Extract a DDD-style ubiquitous language glossary from the current conversation, flagging ambiguities and proposing canonical terms. Saves to UBIQUITOUS_LANGUAGE.md. Use when user wants to define domain terms, build a glossary, harden terminology, create a ubiquitous language, or mentions "domain model" or "DDD".

UI/UX design intelligence. 67 styles, 96 palettes, 57 font pairings, 25 charts, 13 stacks (React, Next.js, Vue, Svelte, SwiftUI, React Native, Flutter, Tailwind, shadcn/ui). Actions: plan, build, create, design, implement, review, fix, improve, optimize, enhance, refactor, check UI/UX code. Projects: website, landing page, dashboard, admin panel, e-commerce, SaaS, portfolio, blog, mobile app, .html, .tsx, .vue, .svelte. Elements: button, modal, navbar, sidebar, card, table, form, chart. Styles: glassmorphism, claymorphism, minimalism, brutalism, neumorphism, bento grid, dark mode, responsive, skeuomorphism, flat design. Topics: color palette, accessibility, animation, layout, typography, font pairing, spacing, hover, shadow, gradient. Integrations: shadcn/ui MCP for component search and examples.

Use when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback

Use only when the user explicitly asks to use Superpowers or the using-superpowers skill. Do not activate automatically at conversation start.

Vercel Agent guidance — AI-powered code review, incident investigation, and SDK installation. Automates PR analysis and anomaly debugging. Use when configuring or understanding Vercel's AI development tools.

Vercel CLI expert guidance. Use when deploying, managing environment variables, linking projects, viewing logs, querying metrics, managing domains, managing feature flags with vercel flags, or interacting with the Vercel platform from the command line.

Vercel Connect expert guidance — securely obtain scoped OAuth tokens for third-party services (Slack, GitHub, MCP servers, OAuth, Snowflake) on behalf of apps or users via Vercel OIDC. Use when wiring up third-party API access, connecting to MCP servers, sending Slack messages, accessing GitHub APIs, receiving webhook events from Slack/Linear/GitHub and forwarding them to your agents and apps, or building eve agent connections.

Vercel Firewall expert guidance — automatic DDoS mitigation, the Vercel WAF (custom rules, IP blocking, managed rulesets, rate limiting), Attack Mode, system bypass, bot management, and the `vercel firewall` CLI. Use when configuring platform-level security, responding to attacks, or staging firewall rules.

Vercel Functions expert guidance — Serverless Functions, Edge Functions, Fluid Compute, streaming, Cron Jobs, and runtime configuration. Use when configuring, debugging, or optimizing server-side code running on Vercel.

React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.

Vercel Sandbox guidance — ephemeral Firecracker microVMs for running untrusted code safely. Supports AI agents, code generation, and experimentation. Use when executing user-generated or AI-generated code in isolation.

Configure and troubleshoot Vercel Services for multiple frontends and backends in one project. Use when composing a polyglot or multi-service application on one Vercel deployment; defining the `services` key, service-targeted rewrites, or service bindings in `vercel.json`; or running all services with `vercel dev`.

Vercel storage expert guidance — Blob, Global Config (formerly Edge Config), and Marketplace storage (Neon Postgres, Upstash Redis). Use when choosing, configuring, or using data storage with Vercel applications.

Full-story verification — infers what the user is building, then verifies the complete flow end-to-end: browser → API → data → response. Triggers on dev server start and 'why isn't this working' signals.

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always

visionOS skills — spatial design (layout ergonomics, eyes-and-hands input, motion comfort, immersion, environments) and widget patterns (mounting styles, glass/paper textures, proximity-aware layouts). Use when designing, building, or reviewing anything for visionOS.

Create visualizations and interactive tools directly in conversation. Proactively use to show how something works; explore 'what happens when', 'what changes', or 'help me understand'; compare or inspect; create simulations, maps, charts, graphs, and mockups. Use standard tools for static scientific figures.

watchOS development guidance including SwiftUI for Watch, Watch Connectivity, complications, and watch-specific UI patterns. Use for watchOS code review, best practices, or Watch app development.

Plan a huge chunk of work — more than one agent session can hold — as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.

Review UI code for Web Interface Guidelines compliance. Use when asked to "review my UI", "check accessibility", "audit design", "review UX", or "check my site against best practices".

Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.

Vercel Workflow SDK expert guidance. Use when building durable workflows, long-running tasks, API routes or agents that need pause/resume, retries, step-based execution, or crash-safe orchestration with Vercel Workflow.

Writing, exploit — assemble raw material into a journey of beats, grounding each term before a beat leans on it.

Writing, explore — mine raw fragments, no structure yet.

Reference for writing and editing skills well — the vocabulary and principles that make a skill predictable.

Use when you have a spec or requirements for a multi-step task, before touching code

Writing, exploit — shape raw material into an article, paragraph by paragraph.

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