convez.ai
SkillProdutividadeIntermediário
nitayk

AI Coding Rules

nitayk/ai-coding-rules

57 skills + 200+ regras + agents para turbinar dev com Cursor/Claude

Ver no GitHub
InicialPluginScore · de 0 a 100

Kit de desenvolvimento com 57 skills, 9 agents, 200+ regras curatorias e hooks para Cursor e Claude Code. Combina as melhores práticas da comunidade (obra/superpowers, anthropics/skills) com workflows battle-tested, funcionando como submódulo git em qualquer projeto.

Instalação

$npx convez-ai nitayk/ai-coding-rules

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 2 semanas
Versão
—
Issues abertas
1
Contribuidores
1

Casos de uso

  • Acelerar desenvolvimento assistido por IA com skills pré-configuradas
  • Aplicar padrões de código consistentes via 200+ regras customizadas
  • Automatizar code review, PR creation e validação de dados com agents
  • Usar hooks de qualidade, segurança e observabilidade

Veredito

Repositório consolidado e maduro que reúne rigorosamente skills, agents e regras de múltiplas fontes confiáveis. Excelente para times que usam Cursor/Claude Code e querem baseline sólido de qualidade; requer familiaridade com git submodules e CLI Go, mas o payoff é alto.

README

A tradução para o português deste README ainda está na fila.

ai-coding-rules

A curated, best-of-breed AI coding toolkit that combines top community skills with battle-tested rules and workflows. Works as a git submodule for any project, instantly supercharging your AI-assisted development.

Supports both Cursor and Claude Code.

What's Inside

CategoryCountSource
Skills57obra/superpowers, anthropics/skills, mcr, Unity, custom
Agents9architect, code-reviewer, data-validator, and more
Commands12brainstorm, create-pr, fix-issue, council, and more
Hooks20+Quality, security, observability, ECC
Cursor Rules200+Custom (Scala, Python, Go, Java, PHP, JS/TS, Swift, Kotlin, Obj-C)
Claude Rules7Custom
CLI Tool Guides13curl, jq, yq, git, docker, kcat, kubectl, aws-cli, gcloud, helm, terraform, ripgrep, gh-cli

Quick Start

Option A: Per-project setup (recommended if you also use work repos)

Cleanest separation -- your personal tools stay in personal projects, zero chance of mixing with org repos.

The shell scripts were replaced by a single Go CLI, acr. Build it once:

# One-time: clone to your home directory and build the CLI
git clone https://github.com/nitayk/ai-coding-rules.git ~/ai-coding-rules
cd ~/ai-coding-rules && go build -o ~/.local/bin/acr ./cli   # ensure ~/.local/bin is on PATH

See cli/README.md for full command docs.

# Link the checkout into a project and deploy skills/rules/agents/commands
acr link ~/projects/my-app
cd ~/projects/my-app && acr sync          # add --copy to copy instead of symlink

# Restart Cursor -- done! 57 skills + 200+ rules in that project.

To update a project later:

cd ~/ai-coding-rules && git pull && acr update   # refresh community sources
cd ~/projects/my-app && acr sync                 # re-deploy

Option B: Git submodule (team/open-source projects)

Version-pinned rules committed to the project repo. acr install adds the submodule, syncs, and installs a post-merge hook that re-syncs after git pull:

git submodule add https://github.com/nitayk/ai-coding-rules.git .cursor/rules/shared
go build -o ~/.local/bin/acr ./.cursor/rules/shared/cli
acr install --target cursor               # or: --target cursor,claude

Global (~/.cursor/) skill install is no longer supported — skills must live in per-project directories.

What goes where (Cursor architecture)

WhatGlobal (~/.cursor/)Project (.cursor/)How
Skills~/.cursor/skills/.cursor/skills/Both work, project takes precedence
Rules (.mdc)Not supported on disk (Settings UI only).cursor/rules/Must be per-project
Agents~/.cursor/agents/.cursor/agents/Both work
Commands~/.cursor/commands/.cursor/commands/Both work

Source: Cursor Skills Docs, Cursor Rules Docs

Skills Inventory

Planning & Design

SkillDescription
brainstormingSocratic design refinement before coding
writing-plansBreak work into 2-5 min bite-sized tasks
executing-plansBatch execution with human checkpoints
prd-generationPRD generation via structured conversation
task-breakdownBreak PRDs into parent tasks and sub-tasks

Development Workflows

SkillDescription
tdd-workflowRed-Green-Refactor cycle
test-driven-developmentRED-GREEN-REFACTOR with anti-patterns reference
git-workflowStable-state commit workflow for complex tasks
pr-workflowFull PR lifecycle (pre-checks, monitoring, merge)
create-prQuick PR creation with template detection
fix-issueFix a GitHub issue end-to-end
setup-local-devPersistent dev server with pm2
finishing-a-development-branchMerge/PR/cleanup decision workflow

Code Quality

SkillDescription
best-practices-enforcementValidates code against universal standards
code-cleanupThree-phase cleanup (AI slop, dead code, anti-patterns)
code-review-excellenceEffective code review practices
code-optimizationPerformance and efficiency optimization
code-migrationCode migration patterns
requesting-code-reviewPre-review checklist
receiving-code-reviewResponding to feedback with rigor
verification-before-completionEvidence before claims
security-auditSecurity vulnerability scanning

Debugging & Testing

SkillDescription
systematic-debugging4-phase root cause process
debug-workflowLog-Reproduce-Fix cycle
webapp-testingPlaywright-based web app testing
test-until-passIterative test-fix loop

Multi-Agent & Parallel

SkillDescription
dispatching-parallel-agentsConcurrent subagent workflows
subagent-driven-developmentFresh subagent per task with two-stage review
multi-agent-branchingFeature-branch isolation for concurrent agents
using-git-worktreesParallel development branches
mass-repo-orchestrationMulti-repo batch operations
councilMulti-perspective decision making

Research & Knowledge

SkillDescription
deep-researchMulti-source research with citations
search-firstResearch before coding
onboard-developerDeveloper onboarding guide
generate-docsAuto-generate documentation
generate-changelogGenerate changelogs from git history
doc-coauthoringCollaborative document creation
address-pr-feedbackSystematic PR feedback handling

Frontend & Documents

SkillDescription
frontend-designProduction-grade frontend interfaces
web-artifacts-builderMulti-component HTML artifacts
agent-browserBrowser automation skill
docxWord document creation, editing, tracked changes
pdfPDF extraction, creation, merge/split, forms
xlsxSpreadsheets with formulas, formatting, analysis
pptxPowerPoint creation and editing
gdocGoogle Docs integration

Meta & Tools

SkillDescription
skill-creatorGuide for creating effective skills
writing-skillsCreate new skills following best practices
mcp-builderGuide for building MCP servers
session-memoryPersistent context across sessions
using-superpowersIntroduction to the skills system
agent-token-optimizationLLM token cost optimization
continuous-learning-v2Instinct-based learning system
strategic-compactContext compaction at logical intervals
prompt-optimizerOptimize LLM prompts
repository-organizationRestructure messy repos

Cursor Rules System

The intelligent Router (ROUTER.mdc) auto-detects your intent and loads the right rules:

  • generic/ - Architecture, code quality, testing, performance, security, debugging, communication, agent behavior
  • backend/ - Scala, Python, Go, Java, PHP patterns
  • frontend/ - JavaScript/TypeScript, Vue, React, accessibility, performance, security
  • mobile/ - Swift, Kotlin, Objective-C, CocoaPods, Gradle
  • tools/ - CLI references for curl, jq, yq, git, docker, kcat, kubectl, aws-cli, gcloud, helm, terraform, ripgrep, gh-cli

Optional Power-Ups

These are full frameworks best used via their own install mechanisms:

  • BMAD-METHOD (36k stars): npx bmad-method install - Full agile AI development with 12+ specialized agents
  • wshobson/agents (28k stars): /plugin marketplace add wshobson/agents - 112 agents, 146 skills, 73 plugins
  • obra/superpowers marketplace: /plugin marketplace add obra/superpowers-marketplace - Plugin system for Claude Code

MCP Configuration

See mcp/example-config.json for a ready-to-use configuration with public MCP servers:

  • context7 - Library documentation lookup
  • playwright - Browser automation
  • puppeteer - Browser testing
  • github - GitHub API integration
  • filesystem, git, shell - Local development tools

Adding More Skills

Skills are just markdown files. You can add more from:

Install with: npx skills add owner/repo --skill "skill-name"

Updating

Pull latest + update community sources

cd ~/ai-coding-rules
git pull
acr update                        # fetch latest from obra/superpowers + anthropics/skills
git add -A && git commit -m "chore: update community skills"

Re-apply to projects

# Linked or submodule project
cd ~/projects/my-app && acr sync

See SOURCES.md for full provenance tracking.

Credits

Built on the shoulders of giants:

License

MIT License - see LICENSE for details.

Skills from external sources retain their original licenses:

  • obra/superpowers: MIT License
  • anthropics/skills: Apache 2.0 (example skills), source-available (document skills)
Ver o original no GitHub

Bom para

Relacionados