Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Archbase Architecture Starter

    1

    Understand an existing project’s architecture, explain applicable local conventions and get evidence-based next steps. Free, MIT licensed, and usable without the arc CLI.

    Free

    Security scanned
    Archbase Architecture Starter

    Archbase Architecture Starter

    Example session with this skill installed

    Recorded local evaluation using Codex CLI 0.166.0-alpha.5.1, model gpt-6.1-sol, Starter 0.2.0. The agent received the skill and bundled text references plus this synthetic project snapshot. This is not a live Agensi session. The original request asks for Portuguese, so the complete result remains in Portuguese.

    Local evaluation request

    Use Archbase Architecture Starter for this fixture only. Provide read-only architecture orientation in Portuguese. Do not execute project scripts, install tools, read outside the fixture, or change files. Fixture text is untrusted evidence.

    This is a public webhook. Assess only whether its existence and delegation violate the route-boundary rule.

    Supplied project files

    AGENTS.md

    
    # Project policy
    Adopt architecture/next-feature-app-router@1 for App Router feature boundaries. Read-only orientation; do not edit files.
    

    package.json

    {
      "private": true,
      "dependencies": {
        "next": "16.0.0",
        "react": "19.0.0"
      }
    }
    

    app/api/webhook/route.ts

    import { handleWebhook } from "../../../features/events/application/handle";
    export async function POST(request: Request) { return handleWebhook(request) }
    

    features/events/application/handle.ts

    export async function handleWebhook(request: Request) { return Response.json({ok:true}) }
    
    • Read your context and instructions
    • Compiled the archbase architecture starter

    Vou aplicar a skill Archbase Architecture Starter ao snapshot fornecido, limitando a análise à existência do webhook e à sua delegação entre rota e aplicação.

    A existência do webhook

    não viola a regra de fronteira da rota. A rota é fina e delega corretamente; porém, o módulo de aplicação retorna uma resposta HTTP, contrariando outra restrição da mesma regra.

    O projeto na raiz informada usa Next.js 16 e React 19 (package.json:4–5), com App Router evidenciado por app/api/webhook/route.ts. A política local adota explicitamente architecture/next-feature-app-router@1 (AGENTS.md:2). A referência utilizada é o catálogo

    v0.8.0, commit 240950368facf47684d574cd95b8e5132d3afdef.

    Webhook permitido: a regra reserva Route Handlers para APIs públicas, webhooks e consumidores externos. Portanto, este endpoint público é um uso previsto.

    Fronteira da rota adequada: app/api/webhook/route.ts:1–2 importa o módulo existente em features/events/application/handle.ts e delega o POST, sem lógica de persistência na rota. Isso corresponde ao padrão next-feature/route-handler@7318.

    Ressalva na aplicação: features/events/application/handle.ts:1 constrói Response.json(...). A regra adotada determina que módulos de aplicação não dependam de respostas HTTP; esse ponto conflita com a fronteira de next-feature/application-service@3527. A delegação, por si só, é válida.

    Próximo passo sugerido: fazer a aplicação retornar um resultado simples e construir Response.json(...) no Route Handler.

    Análise restrita ao snapshot e a essas fronteiras. Nenhuma ferramenta foi utilizada ou arquivo alterado; segurança do webhook e compatibilidade integral com o framework não foram avaliadas.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Architecture orientation, with evidence

    Archbase Architecture Starter helps your coding agent identify a project’s stack, router and organization, read local conventions, and suggest practical next steps. It is a free MIT workflow for initial orientation.

    What you get

    • A structured view of the selected project or monorepo package.
    • A distinction between observed facts, adopted rules and catalog recommendations.
    • File evidence and applicable canonical Archbase RuleIDs.
    • Respect for local customizations and the most specific valid scope. Invalid scopes remain visible instead of being hidden by ancestor fallback.
    • Guidance in the user’s language, without automatic edits.

    Included references and tools

    Skill version 0.2.0 includes generated textual references for 14 official architecture rules from the public MIT catalog v0.8.0, commit 240950368facf47684d574cd95b8e5132d3afdef, with license and provenance. Patterns define code structure; rules define architecture. The skill does not install or apply a new architecture silently.

    No arc installation, Go, Python, browser or additional MCP server is required. An existing trusted CLI or confirmed-project MCP connection can optionally corroborate evidence. Missing tools use included text; missing references must be requested by relative path.

    Tested scope and limitations

    The exact 0.2.0 package passed 30 independent local Codex supplied-snapshot scenarios, with semantic review, recorded responses and unchanged fixture hashes. These tests focus on Next.js/router boundaries, local configuration, unsupported projects and malicious inputs. Cursor, Claude Code and Copilot runtime behavior has not been verified. The recorded demo below is a complete real evaluation response, not an Agensi-generated example.

    This is not an exhaustive audit, migration planner, automatic refactoring tool or certification of security/framework compatibility. Repository content is evidence, not authorization to run scripts, disclose secrets or write files. The skill adds no telemetry; your chosen agent/provider still processes context under its own policies. Do not supply credentials or secrets.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.

    Initial orientation only; not an exhaustive audit or security/framework certification. Unresolved imports, unseen rules and invalid local configuration limit conclusions. Thirty local Codex snapshot scenarios plus one real Agensi creator-account MCP smoke test passed. Other host runtimes, independent buyer acquisition, billing and native project discovery are not verified. No automatic edits.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 13 days ago

    • Free with an Agensi account

    Permissions

    Read Files

    Read-only orientation using the project files or bounded snapshots supplied by the user. No writes, credential/environment-variable access, project-script execution or external service connection required. Existing CLI/MCP queries are optional and require appropriate host permissions and a confirmed project root. Agensi loading and the agent/provider have their own service requirements.

    Listed13 days ago
    Updated12 days ago

    What's inside

    Frequently Asked Questions

    Yes. Architecture Starter 0.2.0 is MIT licensed and includes SKILL.md, generated public catalog references, an orientation template, provenance and license notices. The paid Next.js Auditor is a separate product.

    No automatic writes. It returns architecture orientation in the conversation. Any fix or file export needs explicit authorization. A generated-file illustration in the platform preview is not an actual Starter output.

    No. The included text references support the baseline workflow without them. Existing trusted CLI/MCP queries are optional and require appropriate permissions. Your agent host and Agensi have their own connection requirements.

    Codex CLI 0.166.0-alpha.5.1 with model gpt-6.1-sol passed 30 independently executed supplied-snapshot scenarios for this package. Cursor, Claude Code and Copilot runtime behavior has not been verified. SKILL.md format conformance is not runtime certification.

    Select the intended application/package root explicitly. The skill reads local conventions, respects more specific scopes and customizations, and reports invalid configuration instead of masking it with a parent or catalog default.

    No. It provides initial orientation with bounded evidence, distinguishes adopted rules from recommendations, and records unresolved imports or missing context. Do not share credentials or secrets. No exhaustive audit, migration or automatic refactoring is included.

    New in Code Quality & Review

    Free