aphrody

Develop AI-powered applications using Genkit in Node.js/TypeScript. Use when the user asks about Genkit, AI agents, flows, or tools in JavaScript/TypeScript, or when encountering Genkit errors, validation issues, type errors, or API problems.

Apply Google's design language to a UI. Load whenever the user wants something to look or feel like a Google product, including requests like "apply Google design style", "make it look like a Google product", "Google-style UI", "design.google guidance", "Google design principles", "Google Sans typography", "Product Sans", "Gemini visual language", "the Gemini sparkle / gradient look", "Material You expressiveness", or any question about how Google approaches color, type, motion, and brand across its products (Search, Gemini, Workspace, Pixel). This is the language layer above the Material 3 spec; for exact M3 token values use the sibling skill m3-docs.

Scaffold a self-contained Material Design 3 md-* web component and/or its React wrapper. Use when the user says: "create an M3 component", "scaffold a Material Design 3 <X> component", "build an md-* web component", "make a new Material component", "wrap a component in React", "add md-<x> to the monorepo".

User-invoked fix skill. Auto-correct Material Design 3 violations in a codebase. Use when asked to "fix M3 violations", "auto-correct Material Design issues", "make this M3-compliant", "clean up MUI residue", "fix hardcoded colors", "fix sx on md components", or "fix wrong icon names".

Turn a natural-language design brief into a Material Design 3 React scaffold with @aphrody-code/m3-design. Use when asked to "scaffold a screen from a description", "generate an M3 layout from a brief", "design compiler", "brief to JSX", or "make an M3 prototype/dashboard/deck from this prompt".

Load this skill whenever the user wants to design or build UI that looks Material Design 3 / Material You / Google-style, including requests like "make this look Material", "apply M3 styling", "which color role or token should I use", "design a Google-style component", "is this on-brand for Material", "M3 color system", "how do I use MD3 tokens on the web", "style this with Material Design", "use M3 elevation or shape", "add M3 motion or animation", "pick the right surface level", "make it look like Google Workspace", or any question about MD3 semantics, color roles, type scale, elevation, shape, motion easing, state layers, or window size classes.

Translate and generate documentation for the monorepo with the @aphrody-code/doc-ai CLI (Google Gemini backend, offline fallback). Use when asked to "translate the docs", "generate an API guide for a Lit component", "sync docs to another language", or "run doc-ai".

Answer Material Design 3 specification questions with attribution. Model-invocable and user-invocable. Use when asked "what does the M3 spec say about X", "look up Material Design docs", "official M3 reference for color / type / shape / motion / layout / components", "cite the Material guideline", or "find the m3.material.io page for X".

Generate and apply Material You dynamic color and M3 design tokens at runtime with @aphrody-code/m3-tokens. Use when asked to "theme from a seed color", "Material You runtime", "derive --md-sys-color-* roles", "re-theme live", "dark mode tokens", "convert a MUI theme to M3 tokens", "M3 breakpoints / window size classes", or "expose M3 tokens to Tailwind".

Set up and run @aphrody-code/eslint-plugin-m3 to lint a site that consumes material-web. Use when asked to "lint M3 usage", "set up the m3 eslint/oxlint plugin", "catch sx on md-*", "validate icon names", "enforce M3 color roles", "find leftover MUI props", or "guard a migration with lint".

Apply Material Design 3 motion (transitions, easings, durations, springs) in React with @aphrody-code/m3-motion. Use when asked to "animate with M3 motion", "M3 transition pattern", "fade-through / shared-axis / container transform", "stagger a list", "M3 easing/duration tokens", or "animate an md-* enter/exit".

Audit a file, component, or directory for Material Design 3 conformance. Use when the user says: "check this is M3-compliant", "audit Material Design conformance", "spec-check this component", "spec-check this page", "is this on-spec for M3", "validate my Material UI", "does this follow MD3".

Wire Material Design 3 tokens into Tailwind v4 and shadcn/ui with @aphrody-code/m3-theme and @aphrody-code/m3-tokens. Use when asked to "use M3 with Tailwind", "M3 + shadcn", "Tailwind v4 @theme with --md-sys-* tokens", "style the host of an md-* with Tailwind", "fusion tokens", or "map shadcn color aliases to M3 roles".

Action skill: produces a complete, ready-to-use M3 page or screen scaffold. Invoke when the user says "scaffold an M3 page", "scaffold an M3 screen", "give me a Material Design template", "give me a Material Design layout", "create a dashboard in M3", "create a settings screen in M3", "create a list-detail screen in M3", "create a feed screen in M3", "M3 page boilerplate", or any similar request to generate a concrete screen.

Integrates Material UI with Next.js App and Pages routers using @mui/material-nextjs, Emotion cache providers, next/font, CSS layers with Tailwind/CSS Modules, Link component prop patterns, CSS theme variables SSR notes, and App Router useSearchParams + Suspense. Use when setting up or debugging MUI in a Next.js app.

Chooses the right Material UI styling approach (sx, styled, theme overrides, global CSS) from official MUI guidance. Use when styling @mui/material components, customizing themes, overriding slots, or comparing sx vs styled.

Integrates Material UI with Tailwind CSS v4 using cascade layers (enableCssLayer, @layer order) and documents Tailwind v3 interoperability (preflight, important, injectFirst, portals). Use when combining MUI with Tailwind utilities, slotProps className, or theme token bridges.

Guides Material UI theming and design tokens (createTheme, ThemeProvider, palette, colorSchemes, cssVariables, theme.vars, dark mode, TypeScript augmentation). Use when building or extending a theme, toggling light/dark, or aligning tokens across an app.

Migrate a React + MUI (@mui/material@9 + @mui/icons-material + MUI X) codebase to material-web (@aphrody-code/m3-react + Material Symbols). Use when asked to "migrate MUI to material-web / M3", "port a MUI app", "run the migration codemods", "convert @mui/icons-material to Material Symbols", or "estimate migration effort".

Teaches Claude how to apply Material Design 3 adaptive layout: window size classes, navigation pattern swaps, canonical layouts, grid/margin values, and touch-target rules. Invoke when the user says "make this responsive", "adaptive layout", "what breakpoint", "window size class", "navigation rail vs bar vs drawer", "list-detail layout", "how should this reflow on tablet/desktop", or "M3 layout".

Forced-loop coding mode. Wraps /loop 30s /aphrody-yolo-grind with an explicit perfection oracle — does NOT exit until the codebase meets all objective metrics. Every shell + every agent runs in background to maximize parallelism.

Continuous parallel-grind mode that dispatches parallel YOLO agents per tick to drive open plan/todo items to production-ready as fast as possible. Automatically detects the codebase language, build system, and plan files.

Design, implement, and debug autonomous AI agents and multi-agent systems using the Google Antigravity (AGY) SDK. ACTIVATE this skill when the user wants to create, configure, or orchestrate Google Antigravity agents.

One-session autonomous build of the `aphrody-agent-home` crate (soul / identity / workspace). Dispatches the dedicated agent-home-builder subagent (Opus 4.7) to execute the ENTIRE docs/plans/agent-home.md (AH-0..AH-19) with zero stub and no human in the loop, then verifies, commits in batches, and flips the plan items to done.

Contexte direct des commandes IA / génératif / créatif du CLI aphrody — chat, gemini, agy, agy-loop, antigravity, hermes, notebooklm, image, firefly, memory, design, logo. Use quand il faut savoir ce que fait une de ces commandes, ses flags, ses variables d'env et son feature-gate avant de l'invoquer.

Contexte direct des commandes de reverse engineering, forensics et analyse repo du CLI aphrody — re (triage/strings/sections/auto), forensics, scan, chromium, auto. Use pour savoir ce que fait une de ces commandes, ses sous-commandes, son feature-gate et ses garanties de sécurité avant de l'invoquer.

Contexte direct des commandes système, cycle de vie, réseau et intégration du CLI aphrody — auth, doctor, version, completions, self, oc-onboard/reset/uninstall/pairing/docs, mirror, cros, index, search, dns, a2a, mcp, notify, ide, term. Use pour savoir ce que fait une de ces commandes, ses flags et ses effets (dont les opérations destructives) avant de l'invoquer.

Continuous 2-Claude coordination duel over the A2A file-based protocol (ai.json + .coord JSONL mailbox + HTTP listener). Each iteration writes one envelope into the peer's inbox, runs a code-reviewer pass, and schedules the next tick. Mission target — sustain the duel across hours/days without human handhold.

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.

Orchestrates automated reverse engineering on a binary or directory. Use when you need a complete RE report on a PE/ELF binary without specifying individual passes. Chains triage, strings, Google endpoint extraction, Go detection, and entrypoint disasm in a single command. Pairs with deep-analysis (Ghidra) for decompilation.

Launch the aphrody autopilot loop — Claude Code + Gemini CLI pilot the repo in parallel, infinitely, zero human-in-loop. Use whenever the user says 'lance l'autopilot', 'autopilot', 'duel infini', 'go forever', 'pilote claude et gemini', 'tu pilotes en boucle', 'fais tourner', 'mode auto', 'autonome', or any variant that means 'keep working on aphrody by yourself without me'. Spawns `scripts/autopilot.sh` in the background ; the loop picks the next ⏳ item from `docs/PLAN.md`, dispatches it to Claude (`claude -p`) and Gemini (`aphrody gemini` / resolved `gemini` binary) in parallel each tick, logs NDJSON to `var/log/autopilot.jsonl`, bumps `ai/heartbeat.txt`. Stop with `kill $(cat var/run/autopilot.pid)`.

Canonical Rust 2026 stack chooser — recommends the right crate (+ exact version + alternative rejected + reason) for any domain : HTTP servers, async runtimes, databases, crypto, parsing, CLI, terminal UIs, WASM, GUI, observability, ML. Use this skill *aggressively* — whenever the user asks 'which crate', 'best library for X', 'recommend a stack', 'should I use Y or Z', 'how do I do some-thing in Rust', is about to write or edit any Cargo.toml, mentions adding a dependency, or hesitates between several libraries. Also use proactively before suggesting any crate when discussing architecture, even if the user did not ask explicitly. The recommendations are fact-checked against awesome-rust top-ranked sources, RUSTSEC advisories, OSI license safety (no GPL contamination of Apache-2.0 aphrody), and last-commit recency (active maintenance). Refuses deprecated stacks (openssl 0.9, actix-web 4.x for new code, tokio 0.2, hyper 0.14 raw, native-tls on Linux servers, GPL crypto wrappers).

Color science expert skill with 286K words of reference material covering OKLCH/OKLAB, palette generation, accessibility/contrast, color naming, pigment mixing, and historical color theory.

Fetch current library documentation, API references, and code examples for any developer technology (React, tokio, wgpu, Prisma, Supabase, Express, Tailwind, Django, Spring Boot, …) — even well-known ones. Use this skill whenever the user asks setup or configuration questions ("How do I configure Axum middleware?"), requests code involving libraries ("Write a Prisma query for…"), needs API references ("What are the Supabase auth methods?"), or mentions a specific framework or crate by name. Use even when you think you know the answer — training data may be outdated. Source — upstream `upstash/context7` (MIT), `skills/context7-mcp/SKILL.md`, imported 2026-05-19 and rewired onto the native Rust `mcp__aphrody__context7_*` tools.

Rust CLI replacements for slow GNU coreutils / busybox / PowerShell — same binary on Linux/Windows/macOS. Use whenever picking a shell command, writing a script, or benchmarking. Maps grep→rg, find→fd, sed→sd, ls→eza, cat→bat, du→dust, ps→procs, htop→btm, etc. with installed-state inventory.

Performs focused, depth-first investigation of specific reverse engineering questions through iterative analysis and database improvement. Answers questions like "What does this function do?", "Does this use crypto?", "What's the C2 address?", "Fix types in this function". Makes incremental improvements (renaming, retyping, commenting) to aid understanding. Returns evidence-based answers with new investigation threads. Use after binary-triage (cf. `aphrody re triage`) for investigating specific suspicious areas or when user asks focused questions about binary behavior.

Deploy application to production. TRIGGER when: user says 'deploy', 'ship', 'push to prod', 'release', 'mise en prod', or asks to restart a service. Handles build verification, systemd services, and deployment validation.

Crawl design.google end-to-end via Edge headless (SPA-aware) — discover every /library article + site page, scrape post-hydration DOM via scripts/edge-mass-scrape.ts, then run the design-google-curator agent to assemble docs/DESIGN.md as the canonical aphrody design reference.

MANDATORY first step whenever the user asks ANYTHING about a library, framework, SDK, CLI tool, cloud service, API, configuration syntax, runtime behavior, version migration, or "how do I do X with Y" — single tool call `mcp__aphrody__docs_auto_search` fans out Context7 (community catalog, version-pinned) + Microsoft Learn (official Azure / .NET / Windows / M365 / Power Platform docs + code samples) + Google search IN PARALLEL via `tokio::join!`, returns one fused markdown report. Covers EVERYTHING — Rust crates (tokio, wgpu, reqwest, axum, serde, hyper, rmcp, …), JS frameworks (React, Vue, Svelte, Astro, …), Python libs (FastAPI, Django, polars, pandas, numpy, …), Azure / AWS / GCP services, .NET APIs, Windows kernel + Win32, mobile SDKs (Swift / Kotlin / Flutter), database engines, build tools. Use EVEN WHEN you think you know the answer — training data is months behind release cycles and APIs drift constantly. Skipping this skill = high risk of hallucinated methods, dead URLs, deprecated patterns. Saves wa

Canonical Google / Material Design 3 design authority for aphrody. Use AGGRESSIVELY whenever the user asks anything about Material Design 3, M3 Expressive, Gemini's visual language, Google Sans typography, design tokens, color roles/schemes/HCT, shape & corner-radius scale, elevation, motion, state layers, adaptive layout & breakpoints, accessibility/contrast, or how to theme/build a UI 'the Google way' — and before recommending any colour, type, spacing or component spec. Grounds every answer in the in-repo design package (docs/design/) and the canonical Rust tokens (crates/m3-tokens) + native components (crates/mui-rs-components), distilled from m3.material.io, design.google, developer.android.com and the user's NotebookLM Google-design corpus. Refuses non-Google design systems as the source of truth (no Bootstrap/Ant/Chakra spec) — those are only mentioned as contrast.