StarchyBomb
UserAgency-grade full-stack website skills for coding agents — kills the default AI look. framer-motion motion, shadcn components, Supabase + Vercel deploy. 8 art directions.
Categories
Indexed Skills (22)
shadcn
Manages shadcn components and projects — adding, searching, fixing, debugging, styling, and composing UI, including chat interfaces. Provides project context, component docs, and usage examples. Applies when working with shadcn/ui, component registries, presets, --preset codes, or any project with a components.json file. Also triggers for "shadcn init", "create an app with --preset", or "switch to --preset".
brainstorm-game-design
Brainstorm-mode for designing a browser game's CONCEPT before any code — pillars, core loop, player fantasy, mechanics, scope, win/loss. Use when the user wants to design a new game, rethink an existing one, or figure out "what is this game." Facilitates the brainstorm AND writes every accepted decision into game-design/GDD.md in the user's project so no agent (Claude, Cursor, Antigravity) ever loses the thread. Kills generic "it's a platformer with power-ups" design.
brainstorm-level-design
Brainstorm-mode for designing a browser game's LEVELS and encounters — layout, pacing, difficulty curve, teaching mechanics, critical path, secrets. Use when designing a level, stage, room, arena, or wave, or planning a game's progression. Reads game-design/GDD.md for the game's pillars, then writes each level to game-design/levels/<name>.md so no agent loses the design. Kills flat, arbitrary levels that ignore the game's own mechanics.
browser-engine
Build browser games that are actually PLAYABLE — 2D or 3D — not tech demos. Use for ANY request to build, prototype, or fix a game that runs in a web browser (canvas/WebGL/WebGPU) — top-down, platformer, shooter, puzzle, 3D. Kills the
game-collision
Make browser-game objects ACTUALLY solid — the fix for the
superlazy
Core silent coding mode. Use before writing or modifying ANY code. Minimal code, minimal output, full quality. Always load this first; every other superlazy-* skill assumes this contract.
ultralazy
Maximum-savings loadout — stacks output economy, input/context economy, and tool-output economy into one skill. Use when the goal is squeezing token/credit cost as hard as possible while keeping quality unchanged or better, or when the user says "ultralazy", "save as much as possible", "minimize cost/credits", or wants every lever at once instead of loading superlazy skills piecemeal.
claude-davinci
Make Claude create actual images, including photorealistic ones. Use whenever the user wants any image, picture, photo, art, illustration, logo, icon, wallpaper, poster, banner, avatar, pixel art, generative art, texture, or asks to "draw" / "paint" / "generate an image" — even though Claude "can't generate images". Photo mode drives a real image model (Gemini/OpenAI/free fallback); vector mode paints with SVG/Pillow/numpy.
isekai-backend
Ship a sellable full-stack app, not a frontend mockup. Use whenever a build needs real data, accounts, payments, or a live URL — auth, database schema, Row Level Security, file storage, or deploying to production. Stack is Next.js + Supabase (Postgres/Auth/Storage) + Vercel, all on free tiers. Pair with the isekai skill for the UI and the shadcn skill for components.
isekai
Agency-grade ($10,000+) website building. Use for ANY website, landing page, portfolio, marketing site, or web app UI — whenever the user asks to build, redesign, or improve a web page, even if they don't mention design. Kills the default AI-website look; motion-native (framer-motion scroll animations), image-rich (people don't read), typography-correct, 5th-grader-clear UX.
superlazy-context
Index-then-retrieve for large files, specs, or docs instead of pasting them whole into context. Use when a doc/spec/log would cost hundreds+ tokens to read in full, when the user says /sc (index) or /uc (retrieve), or asks to "save context on this doc" / "load only what's needed". Lossless — retrieval is verbatim, never summarized.
superlazy-debug
Silent root-cause debugging. Use on any bug, test failure, build failure, or unexpected behavior, BEFORE proposing or writing any fix.
superlazy-mcp
Build MCP servers with minimal tokens. Use when creating or modifying an MCP (Model Context Protocol) server integrating any external API or service, in TypeScript or Python.
superlazy-plan
Silent planning and execution for multi-step tasks. Use when a task spans several files/steps, or when a written plan exists and must be executed.
superlazy-repomap
One-shot signature map of a codebase (function/class names + line numbers, no bodies) instead of a Glob/Grep/Read tour. Use when starting work in an unfamiliar or large codebase, before exploring with multiple directory-wide searches, or when the user asks to "map the codebase" / "what's in this repo".
superlazy-review
Silent code review — giving and receiving. Use after completing a task/feature before calling it done, and whenever review feedback arrives, before implementing any suggestion.
superlazy-skill-creator
Create or improve skills with minimal tokens. Use when the user wants a new skill, wants to edit/optimize an existing skill, or wants a workflow captured as a skill.
superlazy-subagents
Silent subagent-driven plan execution. Use when executing an implementation plan with independent tasks and subagent support is available.
superlazy-tdd
Silent test-driven development. Use when implementing any feature or bugfix in a project that has a test setup, before writing implementation code.
superlazy-verify
Evidence before any completion claim. Use immediately before saying done/fixed/passing, before any ✅, and before committing or creating a PR.
superlazy-worktree
Silent isolated workspace setup. Use when starting feature work that needs isolation from the current workspace, or before executing an implementation plan.
supercontexter
Lossless context saving via index + on-demand retrieval (progressive disclosure). Use when a large file/spec/doc would otherwise be pasted whole into context, when the user says /sc (index a file) or /uc (load index card or retrieve sections), or asks to "compress context", "save tokens on this doc", or "load only what's needed". Keeps a small index card in context; the full text stays on disk and exact sections load only when a task needs them. Not a summarizer — retrieval is verbatim.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.