Claude Skills by majinmagros

Baseline cross-project coding conventions for naming, readability, immutability, and code-quality review. Use detailed frontend or backend skills for framework-specific patterns.

Frontend development patterns for React, Next.js, state management, performance optimization, and UI best practices.

Suggests manual context compaction at logical intervals to preserve context through task phases rather than arbitrary auto-compaction.

Project-specific skill template covering architecture, patterns, testing, and deployment guidance.

Quarkusプロジェクト検証ループ：ビルド、静的分析、カバレッジ付きテスト、セキュリティスキャン、ネイティブコンパイル、本番環境またはPR前の差分レビュー。

Project-specific skill template covering architecture, patterns, testing, and deployment guidance.

Unified knowledge base for AI music production combining best references from: - ai-media-generator (Hao0321): prompt engineering for Suno, Veo, Kling, Seedance, etc. - claude-music (AgriciDaniel): ACE-Step 1.5 local generation references - varg-ai (vargHQ): cloud multi-modal (music, video, image, speech) references

Use when building or reviewing production LLM agent applications — applying the 12-Factor Agents methodology (humanlayer) for reliable, scalable, maintainable agentic software. Triggers on "12 factor agents", "12-factor agents", "agente de produção", "LLM app confiável", "build agent", "arquitetura de agente", "por que agentes falham em produção", "agente stateless", "own your context window". Turns the 12 principles into a review checklist and maps each factor to concrete ECC skills.

Use when interacting with 9Router API gateway, managing local/remote AI proxies, fallback chains, auto-routing, and OpenAI-compatible endpoints for chat, image, TTS, embeddings, and web search. Triggers on "9router", "9Router API", "NINEROUTER_URL", "gateway de IA", "auto-fallback".

Use when managing or configuring LLM agent loops experiencing frequent rate limits (429), quota exhaustion, or provider outages, implementing automatic failover and provider chains via 9Router. Triggers on "quedas de ia", "erro 429", "quota esgotada", "fallback automático", "9router fallback".

Use when building systems where independent AI agents built on different frameworks or by different vendors must discover each other, delegate tasks, and share results — the Agent2Agent (A2A) protocol, the open standard for agent-to-agent interoperability. Triggers on "A2A", "Agent2Agent", "agente falando com agente", "interoperabilidade de agentes", "delegar tarefa pra outro agente", "A2A vs MCP", "protocolo entre agentes". Clarifies when to use A2A vs MCP vs native sub-agents, and how to ex...

Use when design, implement, and audit inclusive digital products using WCAG 2.2 Level AA. Triggers on \"accessibility\".

Use when producing ad creatives at volume: research angles, Ogilvy-style copy, Meta Ads Library checks, carousel builds, Meta publish with UTM, and HTML report. Triggers on \"ad creatives\", \"creative factory\", \"carousel ads\", \"publish meta ads\". Non-triggers: copy only with no production publish or report (use marketing-campaign). Outcome: a batch of published creatives with UTM plus a multi-ad HTML report.

Use when operating paid-traffic accounts through an AI agent — audit Google Ads / Meta Ads accounts, propose budget/bid/audience changes, execute mutations via API with human approval, and keep a hypothesis log per client. Triggers on "otimiza minha campanha", "audita a conta de ads", "pausa palavra cara", "escala o conjunto vencedor", "relatório de tráfego pago", "Google Ads via agente", "Meta Ads via agente", "ads operator". Non-triggers: competitor intel only (use analise-concorrentes), ad...

Use when reporting across multiple ad accounts: aggregate spend, CPA/ROAS, and frequency into one comparative HTML view with keyword-DM funnel events. Triggers on \"multi-ads report\", \"CPA ROAS report\", \"compare ad accounts\", \"keyword DM\". Non-triggers: day-to-day operation of a single account (use ads-operator). Outcome: a comparative multi-account HTML report with keyword-DM funnel events.

Use when making a brand cited and recommended by answer engines (ChatGPT, Perplexity, Gemini) via entity consistency, topic depth, third-party mentions, and prompt tracking. Triggers on \"aeo\", \"geo visibility\", \"answer engine\", \"ai visibility\", \"buyer prompts\". Non-triggers: classic keyword ranking only (use seo), campaign copy with no visibility tracking (use marketing-campaign). Outcome: a tracked prompt set with rising mention share plus a prioritized citation action list.

Use when full-stack diagnostic for agent and LLM applications. Audits the 12-layer agent stack for wrapper regression, memory pollution, tool discipline failures, hidden repair loops, and rendering corruption. Produces severity-ranked findings with code evidence. Triggers on \"agent-architecture-audit\", \"agent architecture audit\", \"audit\".

Use quando precisar de automação web em ALTO NÍVEL (objetivo em linguagem natural → navegação autônoma) — Vercel Agent Browser. Diferente do Playwright MCP (baixo nível, determinístico: click, type, navigate, screenshot), o Agent Browser raciocina sobre o objetivo, navega sozinho, lida com variação de layout/responsividade. Triggers: "agent browser", "vercel agent browser", "automação web linguagem natural", "responsividade auto", "navegação autônoma". Use Playwright MCP (`e2e-testing`, `brow...

Use when exposing a product so AI agents can buy or order through a CLI or MCP with typed commands, schemas, and idempotent operations. Triggers on \"agent commerce\", \"CLI for agents\", \"MCP product interface\", \"agent consumable API\". Non-triggers: building a plain MCP server with no ordering or payment semantics (use mcp-server-patterns). Outcome: an agent-readable product interface with commands, schemas, and safe retry rules.

Use when optimizing agent costs via batch mode — 50-75% savings running 24h ahead, pre-warm sandboxes, multi-model allocation. Triggers on "agent batch mode", "batch processing agents", "pre-warm sandboxes", "multi-model cost allocation", "cost savings 50-75%".

Use when head-to-head comparison of coding agents (Claude Code, Aider, Codex, etc.) on custom tasks with pass rate, cost, time, and consistency metrics. Triggers on \"agent-eval\", \"agent eval\", \"eval\".

Use when designing security for LLM agents and agentic applications — protecting them from prompt injection, jailbreaking, data exfiltration, and malicious tool calls. Triggers on "proteger o agente", "prompt injection", "jailbreak", "guard rails", "guardrail", "segurança de agente de IA", "agente recebeu prompt malicioso", "proteger API de agente", "evitar que a IA faça besteira". Covers layered defenses (input intent-checking, output filtering, tool allowlists, scope enforcement, human appr...

Use when design and optimize AI agent action spaces, tool definitions, and observation formatting for higher completion rates. Triggers on \"agent-harness-construction\", \"agent harness construction\", \"construction\".

Use when structured self-debugging workflow for AI agent failures using capture, diagnosis, contained recovery, and introspection reports. Triggers on \"agent-introspection-debugging\", \"agent introspection debugging\", \"debugging\".

Control insider agent misbehavior: sandboxing, trajectory audit and eval separation. Use when running autonomous or multi-agent systems with tools and shared infra. Use quando isolar sandbox por task, auditar trajectory, separar infra de eval, prevenir sabotagem e colusão de agentes.

Use when adding x402 payment execution to agents with budgets and non-custodial wallets. Triggers on \"agent-payment-x402\", \"agent payment x402\", \"x402\".

Use when use after completing any non-trivial task. The agent self-rates its output on 5 axes — accuracy, completeness, clarity, actionability, conciseness — with concrete evidence per criterion. Produces a structured 1-5 scorecard with specific improvem... Triggers on \"agent-self-evaluation\", \"agent self evaluation\", \"evaluation\".

Use when trimming ECC to a project-specific DAILY vs LIBRARY install plan backed by repo evidence. Triggers on \"agent-sort\", \"sort skills\", \"daily vs library\", \"trim install\".

Use when running uncoordinated agent swarms (not orchestrated delegation) with mailbox/thread messaging, file-claim locking, budget caps, bail-out exits, and adversarial verification. Triggers on \"agent swarm\", \"swarm coordination\", \"claim-file\", \"message board\", \"mailbox\", \"deconfliction\", \"bail-out\", \"referee\", \"swarm takeaways\", \"Herder\". Non-triggers: single-orchestrator delegation (use sessoes-orquestradas/team-agent-orchestration), Docker Swarm (use docker-patterns)....

Mount your personal Top-5 agentic benchmarks aligned to your work instead of a global index. Use when choosing models, building a model stack, or comparing performance/cost/speed. Use quando montar top-5 de benchmarks, escolher modelos, comparar custo x performance, fugir de índice global genérico.

Use when operate as an agentic engineer using eval-first execution, decomposition, and cost-aware model routing. Triggers on \"agentic-engineering\", \"agentic engineering\", \"engineering\".

Use when build persistent multi-agent operating systems on Claude Code. Covers kernel architecture, specialist agents, slash commands, file-based memory, scheduled automation, and state management without external databases. Triggers on \"agentic-os\", \"agentic os\".

Use when one repo must serve Claude, Codex, and other harnesses from a single canonical config without duplicated rules. Triggers on \"agnostic repo\", \"same brain\", \"AGENTS.md canonical\", \"claude code and codex\", \"skills symlink\", \"drift audit\". Non-triggers: memory handoff between agents (use unified-memory). Outcome: AGENTS.md as source of truth with thin per-harness imports, linked skills dir, and a passing drift audit.

Use when engineering operating model for teams where AI agents generate a large share of implementation output. Triggers on \"ai-first-engineering\", \"ai first engineering\", \"engineering\".

Use when monitoring AI governance — UN existential risk, harmonized regulations, supply chain red lines, corporate pressure, Mistral sovereignty. Triggers on "ai governance monitor", "un ai regulation", "existential risk ai", "supply chain red lines", "mistral sovereignty".

Use when generating AI images, video, or audio via fal.ai, Kling, Seedance, Suno, Midjourney. Triggers on \"ai-media-generator\", \"ai media generator\", \"gerar imagem\", \"gerar video\"

Use when regression testing strategies for AI-assisted development. Sandbox-mode API testing without database dependencies, automated bug-check workflows, and patterns to catch AI blind spots where the same model writes and reviews code. Triggers on \"ai-regression-testing\", \"ai regression testing\", \"testing\".

Use when building AI tutor systems for personalized learning — adaptive curriculum, multimodal lessons (text/image/audio/video), visual 2D/3D flows (mindmap, quiz, concept city), Feynman technique exercises, spaced repetition flashcards (SM-2/FSRS), progress tracking, multi-tutor domains. Triggers on "tutor ia", "aprender com ia", "aprender qualquer assunto", "fluxo 2d", "fluxo 3d", "cidade de conceitos", "mindmap tutor", "concept city", "currículo adaptativo", "repetição espaçada", "flashcar...

Use when an AI agent must trade stocks/ETFs/crypto through the Alpaca API — paper-first connector setup, scheduled wakeup loops with file handoff, and financial guardrails. Triggers on \"Alpaca\", \"paper trading\", \"stock trader agent\", \"APCA_API_BASE_URL\", \"trading wakeups\", \"alpaca-broker\". Non-triggers: on-chain/crypto-wallet trading (use llm-trading-agent-security), prediction markets (use ito-trade-planner), generic backtesting without a broker. Outcome: a paper-first Alpaca ope...

Use when researching what competitors are running — ads, funnels, offers, hooks, landing pages. Triggers on "analisa concorrentes", "o que os concorrentes estão fazendo", "Facebook Ads Library", "estratégia do concorrente", "pesquisa de mercado", "benchmark". Delivers a structured competitive report.

Use when clean Architecture patterns for Android and Kotlin Multiplatform projects — module structure, dependency rules, UseCases, Repositories, and data layer patterns. Triggers on \"android-clean-architecture\", \"android clean architecture\", \"architecture\".

Use when generates Angular code and provides architectural guidance. Trigger when creating projects, components, or services, or for best practices on reactivity (signals, linkedSignal, resource), forms, dependency injection, routing, SSR, accessibility (ARIA), a... Triggers on \"angular-developer\", \"angular developer\", \"developer\".

Use when working with facts, paths, commands, or citations in this project. Triggers on "verificar", "valida fonte", "anti-alucinação", "conferir", "saber se existe", "tenho certeza". Prevents invented facts: always verify file paths, tool outputs, and claims against the actual system before acting on them.

Build a new API connector or provider by matching the target repo's existing integration pattern exactly. Use when adding one more integration without inventing a second architecture.

Use when rEST API design patterns including resource naming, status codes, pagination, filtering, error responses, versioning, and rate limiting for production APIs. Triggers on \"api-design\", \"api design\", \"design\".

Use when an operator wants chat-based approval loops plus a multi-source morning digest, where scheduled jobs propose in chat and a human approves before execution. Triggers on \"approval loop\", \"morning digest\", \"chat approval\", \"daily briefing\", \"approved execute\", \"CPA ROAS digest\". Non-triggers: unattended scheduled runs with no approval step (use routines), general background agents with no digest (use autonomous-agent-harness). Outcome: gated proposals approved in chat and ex...

Use when capture architectural decisions made during Claude Code sessions as structured ADRs. Auto-detects decision moments, records context, alternatives considered, and rationale. Maintains an ADR log so future developers understand why th... Triggers on \"architecture-decision-records\", \"architecture decision records\", \"records\".

Write articles, guides, blog posts, tutorials, newsletter issues, and other long-form content in a distinctive voice derived from supplied examples or brand guidance. Use when the user wants polished written content longer than a paragraph, especially when voice consistency, structure, and credibility matter.

Use when scaffolding or operating Astro content sites with components, Markdown collections, bulk publishing, Cloudflare deploy, and PageSpeed checks. Triggers on \"astro site\", \"astro deploy\", \"content collections\", \"cloudflare pages\", \"pagespeed\". Non-triggers: answer-engine citation tracking (use aeo-geo-visibility), generic deploy with no Astro involved (use deployment-patterns). Outcome: a shippable Astro site with typed collections, bulk posts, green build, and live Cloudflare ...

Supervise N agents with 1 brain during-run. Use when coordinating 3+ parallel agents, reviewer is overloaded, routine approvals bury exceptions. TRIGGER when: attention bottleneck, exception inbox, batch approvals, escalation path, N:1 review ratio. Gatilhos PT: arquitetura de atenção, inbox de exceções, revisar N agentes, gargalo de supervisão, escalação para humano.