techstack
Default product-domain stack guidance — frontend, backend, desktop, agent, scientific, static sites, and cross-surface contracts. Covers React, Effect Atom, TanStack Router, Effect Platform on Bun, Astro, coss-ui, Tailwind v4, Turborepo, Expo, NativeWind, React Native Reusables, FastAPI, Axum, Electron, CMake/Qt/VTK/ITK, and contracts (Effect Schema, OpenAPI, AsyncAPI). Use when choosing libraries, runtimes, architecture, refactoring stack, or comparing frameworks.
How do I install this agent skill?
npx skills add https://github.com/keejkrej/based-skills --skill techstackIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill provides architectural guidance and code scaffolds for various technology stacks, including React, TypeScript, Python, and Rust. It promotes security-conscious practices such as runtime validation of untrusted input, the use of official package registries, and standard linting tools.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
Techstack
Global rules
- Effect is the TS domain/IO backbone — structured errors, DI, resource safety, retries — not ad hoc Promises
- One primary server stack per service
- Repo conventions beat greenfield defaults when they conflict
- Typed boundaries at every seam — decode/validate at runtime on untrusted input
- Greenfield installs: seven-day minimum registry release age — see domains/security.md
- TS relative imports (all apps, packages, routes, components): always extensionless —
from './module', neverfrom './module.js'orfrom './module.ts' - Effect implementation depth in Effect repos →
effect-tsskill
Defaults at a glance
Enough to choose libraries and direction; read the linked file before implementing details.
Frontend
- Web SPA: React + Bun + Tailwind v4 + coss-ui + Effect Atom + TanStack Router
- Client IO in Effect programs; state in atoms — not raw
fetchin components - Mobile: React Native + Expo + NativeWind + React Native Reusables (+ Expo Router when file-based routing fits)
- Monorepo (SPA frontend + Rust/Python backend): Bun workspaces + Vite+ tasks (
vp run);contracts+client+utils+ui-headless+ui+web-app - Monorepo (Next.js / fullstack Vercel): Bun workspaces + Turborepo (
turbo run) - Build toolchain, lint/format, component install →
toolingskill - → domains/frontend.md
Backend
- TS API:
@effect/platform(HttpApiorHttpRouter) on Bun — not a separate router framework - Python API/scripts: FastAPI + Pydantic v2 + uv
- Rust API: Axum + serde + utoipa
- WebSockets in Effect services: Effect Platform
Socket - Prototype in Python; harden/distribute in Rust
- → domains/backend.md
Desktop
- UI-first: Electron bundling the same web stack
- Separate process: companion backend on
localhostover WebSocket; preload/IPC only for OS APIs - Native UI: egui (Rust tooling/sim) or Qt 6 (industrial widgets) — not Electron when native perf is the point
- → domains/desktop.md
Agent
- Durable agentic apps on Vercel: Next.js + Eve (filesystem-first) + json-render (generative UI) + agent-browser + wterm/just-bash + Vercel Workflow/Sandbox/Drop
- Models via Ollama (OpenAI-compatible):
http://127.0.0.1:11434/v1using AI SDKopenaiprovider; cloud models as{model}:cloud - Agent is a directory under
agent/: instructions, tools, skills, channels, connections, sandbox, subagents, schedules - Generative UI via constrained json-render catalogs
- Sandbox untrusted code; use
needsApprovalfor high-impact actions - Prototypes via Vercel Drop; move to git for iteration
- → domains/agent.md
Static sites
- Content-heavy sites (blogs, docs, marketing): Astro + static output + Tailwind v4 + Bun
- Islands for client interactivity; avoid SPA stack unless the site is actually an app
- Content Collections for typed structured content; MDX for rich layouts
- → domains/static.md
Scientific
- C++ imaging: CMake + Qt 6 + VTK + ITK
- Rust numerics:
ndarrayecosystem — not hand-rolled NumPy ports - → domains/scientific.md
Contracts
- TS ↔ TS: shared Effect
Schema+HttpApiin monorepo — no codegen between app and API - TS ↔ Python: FastAPI OpenAPI →
openapi-typescript; WebSocket → AsyncAPI - TS ↔ Rust:
utoipaOpenAPI →openapi-typescript; WebSocket → AsyncAPI orasyncapi-rust - IPC (Electron/Tauri): Specta or pydantic-to-typescript
- → domains/contracts.md
Pick by domain
| Domain | Read |
|---|---|
| User-facing SPA or mobile UI | domains/frontend.md |
| API, service, script, or data job | domains/backend.md |
| Desktop or native app | domains/desktop.md |
Agentic apps (Eve + json-render + agent-browser + wterm + Vercel primitives; models via Ollama at 127.0.0.1:11434/v1 with AI SDK openai + {model}:cloud) | domains/agent.md |
| Static content, blogs, docs, or marketing sites | domains/static.md |
| Imaging, numerics, or scientific compute | domains/scientific.md |
| UI/service boundary or cross-language types | domains/contracts.md |
| Supply chain, install hardening, release-age gating | domains/security.md |
| Development tooling, lint/format, CI, scaffolds | tooling skill |
Scaffolds
Greenfield starter structures live in tooling/scaffolds/. Read tooling/scaffolds/README.md before generating files.
Compositions
- Desktop UI + local backend → domains/desktop.md + domains/backend.md + domains/contracts.md
- Full agentic stack (Next.js + Eve + json-render + agent-browser + wterm + Vercel + Ollama via AI SDK openai): domains/agent.md
- New repo setup → domain file(s) +
toolingskill (tooling/domains/dev.md,tooling/scaffolds/) + domains/security.md
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/keejkrej/based-skills/techstack">View techstack on skillZs</a>