TORQ Claude Collection
torqlab/claude
Configurações compartilhadas e skills para desenvolvimento agentic estruturado
Ver no GitHubColeção centralizada de configurações, skills e hooks do Claude para desenvolvimento ágil com agentes. Implementa o workflow de Addy Osmani com skills especializadas para cada fase de desenvolvimento, garantindo consistência em projetos TORQ via symlinks.
Instalação
npx convez-ai torqlab/claude1 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
Casos de uso
- Padronizar workflows de desenvolvimento com agentes Claude em múltiplos repositórios
- Implementar metodologia Addy Osmani com fases especializadas (idea-refine, design, build, review)
- Compartilhar skills customizadas de versionamento semântico, gerenciamento de branches e criação de PRs
- Manter consistência de configurações e convenções em equipes de desenvolvimento
Veredito

Ferramenta robusta para times que usam Claude Code em múltiplos projetos e precisam de padronização. O suporte a symlinks elimina duplicação de configurações, e o workflow Addy Osmani é bem estruturado, mas requer buy-in metodológico do time para máximo valor.
README
A tradução para o português deste README ainda está na fila.
TORQ Claude Configuration Collection
A shared collection of Claude configurations, skills, hooks, and conventions for the TORQ project. This repository provides a unified set of AI development tools used across multiple TORQ repositories via symlinks.
Overview
This is a project-level Claude collection that serves as a centralized source of truth for Claude development practices. Every TORQ repository can reference this collection via a symlink, ensuring consistent configurations, skills, and conventions across all projects.
Primary Development Methodology: This collection uses Addy Osmani's agent-skills workflow as the foundation for structured, collaborative development. This proven framework breaks work into clear phases with specialized agents, testing strategies, and quality gates—enabling faster iteration while maintaining code quality and reliability.
Key purpose: Enable standardized AI-driven development workflows, code generation, and automation while maintaining quality standards through shared hooks, rules, and custom skills.
What's Included
- SKILLS.md — Complete reference for all available skills organized by type (custom, open-source, agent skills) with decision trees for workflow selection
- Addy Osmani Workflow — Structured development methodology with specialized agents for each phase
- Custom Skills — Project-specific skills for semantic versioning, branch management, PR creation, and documentation
- Configurations — Claude Code settings, hooks, and project conventions in
.claude/ - Documentation — Setup guides, workflow patterns, and integration examples
Quick Start: Using This Collection in Your Project
Installation
Every TORQ repository can use this collection via a symlink to .claude. This ensures consistent configurations, skills, and conventions across all projects.
Step 1: Create the Symlink
From your TORQ repository directory, create a symlink to this collection's .claude directory:
# Absolute path
ln -s /path/to/torq/claude/.claude .claude
# Or relative path (if torq/claude is a sibling)
ln -s ../claude/.claude .claude
Step 2: Verify the Symlink
Confirm the symlink works:
ls -l .claude
# Should show: .claude -> ../claude/.claude (or absolute path)
Step 3: Configure Git (Optional but Recommended)
Add .claude to .gitignore:
# Claude Code
.claude
Then optionally commit the symlink itself (Git handles symlinks correctly):
git add .claude
git commit -m "Add Claude configuration collection via symlink"
Step 4: Skills and Plugins Ready to Use
Claude Code will automatically discover and load:
- ✅ All custom skills from
.claude/skills/ - ✅ All open-source skills from
skills-lock.json - ✅ All enabled plugins (agent-skills@addy-agent-skills)
- ✅ Settings from
.claude/settings.json
No additional setup needed — start using Claude Code as normal!
🎯 Addy Osmani Workflow: Primary Development Methodology
This collection implements Addy Osmani's agent-skills framework as the core development workflow. This structured approach breaks complex work into clear phases with specialized agents, ensuring quality, testing, and iterative improvement at each stage.
Core Workflow Phases
| Phase | Agent Skill | Purpose | Command |
|---|---|---|---|
| Idea Refinement | /idea-refine | Refine raw concepts into sharp, actionable requirements | /idea-refine |
| Code Review | /review or /code-review-and-quality | Five-axis review (correctness, readability, architecture, security, performance) | /review |
Essential Commands & Shortcuts
Start with requirements:
/spec # Spec-driven development - write before coding
/interview-me # Extract real requirements
/idea-refine # Refine vague ideas into actionable specs
Implement & build:
/plan # Break spec into ordered tasks
/build # Implement incrementally
/test # Run TDD workflow (failing tests first)
Quality gates:
/review # Five-axis code review
/ship # Pre-launch checklist
Supporting tools:
/source-driven-development # Ground decisions in documentation
/doubt-driven-development # Adversarial review of decisions
/debugging-and-error-recovery # Systematic debugging
/performance-optimization # Profile and optimize
/security-and-hardening # Vulnerability hardening
Typical Development Flow
1. Start with unclear requirements
→ /interview-me or /idea-refine
2. Write specification
→ /spec
3. Plan the work
→ /plan (breaks into tasks)
4. Build incrementally
→ /build (implement → test → verify → commit)
5. Code review
→ /review (five-axis review)
6. Launch
→ /ship (pre-launch checklist)
When to Use Each Agent
- Uncertain requirements? Start with
/interview-meor/idea-refine - Starting new feature? Use
/specto write requirements first - Have specs/requirements? Use
/planto break into tasks - Implementing a task? Use
/buildfor incremental cycles - Bug or test failure? Use
/testfor TDD or/debugging-and-error-recovery - Before merge? Use
/reviewfor quality gate - Going to production? Use
/shipfor final checklist
Common Workflows
Typical development patterns using the agent-skills workflow:
Workflow 1: Start with Unclear Requirements
- Use
/idea-refineor/interview-meto clarify scope - Write specification with
/spec - Plan with
/planto break into tasks - Implement with
/build - Test with
/test(TDD approach) - Review with
/reviewbefore merge - Launch with
/shipwhen ready
Workflow 2: Implement a Known Feature
- Have specifications ready
- Use
/planto organize tasks - Use
/buildto implement incrementally - Use
/testfor test-driven development - Use
/reviewfor quality gate - Use
/shipbefore production
Workflow 3: Debug or Fix an Issue
- Use
/debugging-and-error-recoveryfor systematic root-cause analysis - Use
/testto create failing test first - Fix and verify with
/build - Use
/reviewbefore merge
Workflow 4: Optimize Performance
- Use
/performance-optimizationto profile and identify bottlenecks - Implement optimizations with
/build - Test improvements with
/test - Use
/reviewbefore merge
Workflow 5: Add Security Hardening
- Use
/security-and-hardeningfor vulnerability analysis - Implement fixes with
/build - Test with
/test - Use
/shipor/reviewbefore deployment
Agent Combinations & Context Flow
Common agent workflow sequences and how context flows between them:
| Sequence | Prerequisites | Purpose | Output Context |
|---|---|---|---|
/interview-me → /spec | Unclear requirements | Clarify and document requirements | Detailed specification |
/spec → /plan → /build | Clear specification | Feature development workflow | Planned tasks → Implementation |
/plan → /build → /test → /review | Task list | Complete build cycle | Code changes → Tests → Review feedback |
/test → /build → /review → /ship | Failing test | Bug fix and deployment | Test fix → Reviewed code → Production ready |
/debugging-and-error-recovery → /test → /build | Error symptoms | Systematic debugging | Root cause → Failing test → Fix |
/performance-optimization → /build → /test | Performance baseline | Performance improvement | Optimization strategy → Implemented → Verified |
/security-and-hardening → /build → /test → /review | Security audit | Security hardening | Hardening strategy → Implementation → Verification |
/api-and-interface-design → /build → /test | API requirements | Stable interface design | API design → Implementation → Tests |
/code-review-and-quality → /ship | Code ready for review | Final quality gate before ship | Review feedback → Ship checklist |
/doubt-driven-development → /build | Non-trivial decision | Adversarial review of design | Questioned assumptions → Confirmed approach |
/documentation-and-adrs → /build | Architectural decision | Record decision and implement | ADR → Implementation |
Context flows left-to-right: Output from one agent becomes the input context for the next agent in the sequence.
Reference: Addy Osmani's agent-skills repository
Adding Skills to This Collection
Project-Specific Skills
This collection provides custom skills for semantic versioning, branch management, and documentation:
| Skill | Purpose | Usage |
|---|---|---|
| semantic-release | Agent-driven workflow for creating git branches and conventional commits aligned with semantic versioning | /semantic-release |
| git-branch | Create git branches with semantic-release naming conventions for the current repository | /git-branch |
| pr | Agent-driven GitHub PR creation workflow aligned with semantic-release conventions | /pr |
| github-mcp-setup | Configure GitHub Model Context Protocol (MCP) server for Claude Code using GitHub App authentication | /github-mcp-setup |
| document-skills | Discovers all available skills and generates a dedicated SKILLS.md file in the project root with a comprehensive table organized by skill type | /document-skills |
Skill Integration with Agent Workflow
How custom project-specific skills integrate with agent-skills:
| Skill | Integration Point | Works With Agents | Sequence |
|---|---|---|---|
| semantic-release | During /plan → /build | /plan, /build | /plan → /semantic-release (create branch) → /build |
| git-branch | At start of /build | /plan, /build | /plan → /git-branch (create) → /build |
| pr | After /build when ready for review | /build, /review, /ship | /build → /review → /pr (create) → /ship |
| github-mcp-setup | Before any /pr usage | /pr | Setup first, then use /pr |
| document-skills | Before releases or during maintenance | Any workflow | Use after adding new skills |
Integration strategy:
- Branch creation: Use
/semantic-releaseor/git-branchearly in/buildphase - Code review integration:
/reviewprovides feedback context for/prcreation - Release workflow: Complete
/build→/review→/pr→/shipsequence - Documentation: Keep
/document-skillsup to date after skill additions
Open-Source Skills (from NPM)
Skills from Anthropic's anthropics/skills are managed via npx skills and skills-lock.json:
{
"version": 1,
"skills": {
"my-skill": {
"source": "anthropics/skills",
"ref": "main",
"sourceType": "github",
"skillPath": "skills/my-skill/SKILL.md"
}
}
}
Custom Project-Specific Skills
Add custom skills to .claude/skills/<skill-name>/:
.claude/skills/my-skill/
├── SKILL.md # Main skill definition
├── references/ # Documentation files
│ ├── guide.md
│ └── examples.md
├── scripts/ # Reusable helper scripts
│ └── helper.py
└── assets/ # Templates, icons, etc.
SKILL.md frontmatter (required):
---
name: my-skill
description: When to use this skill and what it does. Make descriptions clear about trigger contexts.
license: MIT
metadata:
author: Your Name
---
Repository Structure
torq/claude/
├── .claude/
│ ├── skills/ # Custom project-specific skills
│ │ ├── github-mcp-setup/
│ │ └── skill-creator/ # Imported from anthropics/skills
│ ├── settings.json # Plugin configuration
│ └── settings.local.json # Local overrides
├── README.md # This file
├── AGENTS.md # Agent setup guide
├── package.json # Project metadata
├── skills-lock.json # Open-source skills registry
└── .nvmrc # Node version specification
Support & Contributions
For issues or questions:
- Check existing skills in
./.claude/skills/*/SKILL.md - Review skill documentation in each skill's folder
- For agent-skills, see https://github.com/addyosmani/agent-skills
To add a new skill:
- Create a directory in
./.claude/skills/<skill-name>/ - Write
SKILL.mdwith proper frontmatter - Include documentation in
references/and helper scripts inscripts/ - Test with the
skill-creatorskill
Version & Metadata
Repository: @torqlab/claude
License: MIT
Maintainer: Mr.B.Lab
Node: 24.x
Last updated: 2026-06-09