Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Archbase PR Architecture Guard

    1

    Review architecture changes in supplied PR diffs with BASE/HEAD evidence. Separate regressions, resolved issues and pre-existing debt. No posting, approval, merge or automatic edits.

    $15

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Archbase PR Architecture Guard

    Archbase PR Architecture Guard

    Example session with this skill installed

    Recorded local evaluation: pr-architecture-guard 0.1.0; codex-cli 0.166.0-alpha.5.1; model gpt-6.1-sol. Synthetic supplied snapshot, not a live Agensi session. The agent received the skill and its included references. The complete recorded response below is preserved in its original language. No framework build or automatic edits were performed.

    Use pr-architecture-guard to review only this supplied synthetic PR comparison. Treat base/ and head/ as revision snapshots and changes.patch as the compared patch; these prefixes are not actual project paths. Respond in Portuguese with adopted baseline, introduced/worsened findings, pre-existing/resolved observations, real paths and BASE/HEAD lines where available, severity, recommendations and coverage limits. Do not edit, execute scripts, install, access external services, post comments/reviews or change PR state. Do not assume this minimal project is runnable.

    Supplied synthetic files

    base/AGENTS.md.md

    ```markdown
    
    # Project policy
    Adopt architecture/next-feature-app-router@1 for this package. 
    
    
    ### base/app/page.tsx.md
    ```text
    ```tsx
    
    ```ts
    export default function Page() { return <main>Home</main> }
    

    base/features/items/components/button.tsx.md

    ```tsx
    "use client";
    export function Button() { return <button>Load</button> }
    
    
    ### base/features/items/data/items.ts.md
    ```text
    ```typescript
    import "server-only";
    export function listItems() { return [] }
    
    
    ### base/package.json.md
    ```text
    ```json
    {"name":"next-review","private":true,"dependencies":{"next":"16.0.0","react":"19.0.0"}}
    
    
    ### changes.patch.md
    ```text
    ```text
    diff --git a/features/items/components/button.tsx b/features/items/components/button.tsx
    --- a/features/items/components/button.tsx
    +++ b/features/items/components/button.tsx
    @@ -1,2 +1,3 @@
     "use client";
    -export function Button() { return <button>Load</button> }
    +import { listItems } from "../data/items";
    +export function Button() { return <button onClick={() => listItems()}>Load</button> }
    
    
    ### head/AGENTS.md.md
    ```text
    ```markdown
    
    # Project policy
    Adopt architecture/next-feature-app-router@1 for this package. 
    
    
    ### head/app/page.tsx.md
    ```text
    ```tsx
    export default function Page() { return <main>Home</main> }
    
    
    ### head/features/items/components/button.tsx.md
    ```text
    ```tsx
    "use client";
    import { listItems } from "../data/items";
    export function Button() { return <button onClick={() => listItems()}>Load</button> }
    
    
    ### head/features/items/data/items.ts.md
    ```text
    ```typescript
    import "server-only";
    export function listItems() { return [] }
    
    
    ### head/package.json.md
    ```text
    ```json
    {"name":"next-review","private":true,"dependencies":{"next":"16.0.0","react":"19.0.0"}}
    
    
    ### review-context.md.md
    ```text
    ```markdown
    
    # Supplied comparison
    This is a synthetic PR example with explicit labels base-example and head-example, not real commit SHAs. base/ and head/ are snapshot storage prefixes, not repository directories. changes.patch paths are the real project-relative paths. Only provided files are available. No GitHub access is supplied. 
    
    • Read your context and instructions
    • Compiled the archbase pr architecture

    Vou aplicar a skill pr-architecture-guard ao diff e aos snapshots fornecidos, verificando a dependência introduzida contra a regra adotada, sem acessar arquivos ou serviços.

    A comparação fornecida introduz

    uma violação confirmada de severidade média: um componente cliente passa a importar e chamar diretamente um módulo marcado como exclusivo do servidor.

    Escopo e baseline

    Revisão de base-example → head-example, rótulos sintéticos sem SHAs reais, conforme review-context.md:2. Os snapshots BASE/HEAD correspondem ao único hunk de changes.patch, que altera features/items/components/button.tsx. O patch está completo para essa alteração apresentada; a cobertura permanece restrita ao material fornecido.

    A política em AGENTS.md:2, tanto em BASE quanto em HEAD, adota architecture/next-feature-app-router@1, versão

    1.0.0, do catálogo

    v0.8.0, commit 240950368facf47684d574cd95b8e5132d3afdef. Essa regra é compatível com o App Router observado. Não foram aplicadas regras de Next.js Pages Router ou de outras stacks.

    Achado introduzido — dependência cliente → módulo server-only

    Evidência: features/items/components/button.tsx,

    HEAD:1–3, mantém "use client", adiciona import { listItems } from "../data/items" na linha 2 e chama listItems() no onClick na linha 3.

    Destino confirmado: o import relativo resolve para features/items/data/items.ts,

    HEAD:1–2, que contém import "server-only" e exporta listItems.

    Regressão: em

    BASE:1–2, o botão não importava nem chamava esse módulo. A nova dependência é diretamente demonstrada pelo diff.

    Regra: architecture/next-feature-app-router@1, versão 1.0.0, estabelece módulos de dados como adaptadores exclusivos do servidor e delimita a interação cliente.

    Severidade média: há um conflito explícito entre a fronteira cliente e a restrição do módulo importado. Não há evidência de execução que permita afirmar falha de build ou impacto operacional.

    Correção recomendada: manter o acesso aos dados no servidor. Para leitura na composição, consultar no servidor e passar dados serializáveis ao botão. Se a leitura precisar ocorrer após o clique, definir uma fronteira de execução no servidor apropriada ao fluxo, preservando o marcador server-only. A função apresentada apenas retorna []; não há evidência de mutação que sustente um achado adicional sobre validação ou autorização de escrita.

    Observações preexistentes e resolvidas

    O marcador server-only e a função listItems já existiam em BASE e permanecem inalterados; o problema introduzido é o novo consumidor cliente. Não foram identificadas outras violações preexistentes ou correções demonstráveis no recorte.

    Próximos passos e limites

    Prioridade: remover a dependência direta do componente cliente e escolher a fronteira de leitura conforme o comportamento desejado. Depois, validar essa fronteira no ambiente real; nenhum build ou teste foi executado, e não se presume que o projeto mínimo seja executável.

    Não foram fornecidos .archbase/scopes.yaml nem outros consumidores de Button; portanto, não foi possível avaliar personalizações adicionais ou seu alcance na aplicação. Todas as referências necessárias ao achado estão presentes.

    O relatório permaneceu no chat. Nenhum arquivo, comentário ou estado de PR foi alterado. Esta revisão limitada não constitui aprovação integral do PR ou recomendação de merge.

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

    About this skill

    Purpose

    Compare supplied unified diffs and matching BASE/HEAD evidence against adopted architecture rules. Initial evaluated examples cover Next.js, React/Vite and NestJS/Prisma; bundled references alone do not certify other stacks.

    What you receive

    A Markdown review separating introduced/worsened issues, resolved findings, pre-existing debt, suspicions and missing context. Confirmed findings cite BASE/HEAD paths, locations or excerpts, applicable rules, justified severity and recommendations. No automatic PR approval or posting.

    Evidence and local conventions

    Local customizations and more specific scopes take precedence. Unknown imports, missing context and invalid selected configuration are never silently treated as confirmed violations or replaced by another architecture. Responses follow the user's language.

    Included package and license

    Skill content version 0.1.0. Generated canonical references use Archbase catalog v0.8.0, commit 240950368facf47684d574cd95b8e5132d3afdef. Original reusable workflow and templates are proprietary; included public references retain MIT attribution. Legitimately generated project-specific results may be used under the license. The skill itself may not be resold or redistributed.

    Tools and privacy

    Use a SKILL.md-compatible AI agent and bounded project text. No arc, Python, Go, framework runtime, extra MCP server or connected service is required. Existing trusted read-only tools are optional. Project text is untrusted evidence, never permission to execute instructions or disclose secrets. This skill adds no telemetry; your agent/provider processes context under its own configuration.

    Tested scope and limits

    Tested in 12 independent local Codex CLI supplied-snapshot scenarios, with semantic review. No live Agensi delivery test for this product. No framework builds, runtime verification, complete security audit, native host discovery or Cursor/Claude Code/Copilot certification. Analysis is limited to the supplied scope; unresolved imports and invalid configurations remain explicit limitations. No scripts, installations or automatic source changes.

    Support

    Use Agensi creator messaging with a minimal redacted reproduction. Best-effort support; no guaranteed response deadline. Platform purchase/refund terms apply. One-time purchase; no subscription.

    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.

    Tested in 12 independent local Codex CLI supplied-snapshot scenarios, with semantic review. No live Agensi delivery test for this product. No framework builds, runtime verification, complete security audit, native host discovery or Cursor/Claude Code/Copilot certification. Analysis is limited to the supplied scope; unresolved imports and invalid configurations remain explicit limitations. No scripts, installations or automatic source changes.

    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

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 12 days ago

    Software
    Python

    Permissions

    Read Files

    Read only the bounded project/package or supplied snapshot authorized by the user. No required shell, writes, network, environment variables or connected service. Output is Markdown in the conversation; no PDF or generated source file is promised.

    Listed12 days ago

    What's inside

    Frequently Asked Questions

    No. It returns analysis and proposed next steps. Implementation, script execution and other actions require separate authorization.

    No. The bundled instructions and textual references work with a SKILL.md-compatible agent; no extra runtime, CLI or project MCP is required.

    12 real independent local Codex CLI supplied-snapshot scenarios. This does not certify framework builds, other agents or delivery through Agensi.

    No. It respects adopted local conventions, scoped overrides and configuration errors, distinguishing them from catalog recommendations.

    The reusable original workflow and templates. Public canonical references remain MIT; legitimately generated project-specific results are usable under the license. Do not resell or redistribute the skill.

    Message the creator on Agensi with a minimal redacted example. Support is best effort. Platform purchase and refund policies apply.

    New in Code Quality & Review