Claude Skills Bank
AlexHandy1/claude-skills-bank
Banco de skills agenticos e workflows para Claude Code
Ver no GitHubColeção pessoal de skills e workflows agenticos para Claude Code, incluindo automação de planejamento, resumos de sessão, testes TDD e automação de navegador. Combina skills customizados com padrões reutilizáveis de outras fontes consolidadas em um repositório único.
Instalação
npx convez-ai AlexHandy1/claude-skills-bank1 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
- Automatizar recapitulação de planos e resumos de sessão
- Estruturar planejamento diário com rastreamento de tarefas
- Implementar workflow TDD com slices verticais
- Automatizar testes comportamentais e padrões de factory
Veredito

Repositório prático que agrupa skills testados para Claude Code em um único ponto de instalação. Útil para quem já trabalha com agentes Claude e quer ganhar velocidade em planejamento, documentação e testes sem reinventar a roda.
README
A tradução para o português deste README ainda está na fila.
AI Agent Knowledge Bank
A personal collection of agentic skills, workflows, and resources I use — a mix of custom ones I've built and links to other resources.
The skills and scaffolding here are primarily designed around Claude Code, though the patterns and ideas are transferable to other agentic setups.
Install
A subset of the skills and workflows I currently use as outlined in the CLAUDE.md file are available to install into any project. Clone this repo, then from your project root run:
mkdir -p .claude && cp -r ~/ai-agent-knowledge-bank/skills .claude/ && cp ~/ai-agent-knowledge-bank/CLAUDE.md .
Adjust the path if you cloned the repo somewhere other than
~.
This copies all skills into your project's .claude/skills/ directory (so they're available in cloud/autonomous sessions) and drops CLAUDE.md at your project root.
I'm still exploring the best way to package and share this as a proper Claude Code plugin — will update once the workflow has stabilised.
Install contents
| Skill | Trigger | Description | Source |
|---|---|---|---|
| plan-recap | /plan-recap | Reads planning docs and produces a concise session briefing | Custom |
| summarise-session | /summarise-session | Writes a structured WORK_SUMMARY_DDMMYY.md from the current session | Custom |
| daily-summary | /daily-summary | Pulls Granola meeting notes and writes a dated daily summary with action items | Custom |
| daily-plan | /daily-plan | Builds the next day's to_do_DDMMYYYY.md from unfinished tasks and meeting actions | Custom |
| grill-me | /grill-me | Stress-tests a plan by interviewing you until reaching shared understanding | Matt Pocock |
| tdd | /tdd | Red-green-refactor TDD workflow with vertical slices | Matt Pocock |
| testing | /testing | Test patterns, factories, and anti-patterns for behaviour-driven tests | Gidi Morris |
| agent-browser | /agent-browser | Browser automation CLI for AI agents via CDP | Vercel Labs |
| documentation-and-adrs | /documentation-and-adrs | Records architectural decisions and ensures README exists | Gidi Morris |
Custom skills
| Skill | Trigger | Description |
|---|---|---|
| plan-recap | /plan-recap | Reads planning and work summary docs from the current project and produces a concise session briefing |
| summarise-session | /summarise-session | Reviews the current conversation and writes a structured summary to a WORK_SUMMARY_DDMMYY.md file |
| daily-summary | /daily-summary | Pulls today's Granola meeting notes and writes a dated daily summary with consolidated action items |
| daily-plan | /daily-plan | Builds the next working day's to_do_DDMMYYYY.md by carrying over unfinished tasks and pulling in meeting actions |
Skills from others
| Skill | Source | Description |
|---|---|---|
| agent-browser | skills.sh | Browser automation CLI for AI agents — fast, persistent browser control with accessibility-tree snapshots and session management |
| xp-and-lean-agent-specification | GitHub | Gidi Morris's XP and lean principles applied to agentic workflows — skills and specifications for structured, disciplined AI-assisted development |
| skills | GitHub | Matt Pocock's collection of Claude Code skills |
| gstack | GitHub | Garry Tan's personal AI development stack and tooling setup |
| gbrain | GitHub | Garry Tan's gbrain — an architectural pattern for agentic memory and reasoning worth exploring |
Plugins from others
Claude Code plugins are a way to package multiple skills, agents, and workflows together into a single installable unit — useful for shipping opinionated setups for a given domain or workflow.
| Plugin | Description |
|---|---|
| frontend-design | Create distinctive, production-grade frontend interfaces with high design quality |
| feature-dev | End-to-end feature development workflow — from spec to working, tested code |
I'm exploring how to package opinionated plugin packs for specific workflows (e.g. data analysis) — bringing together the right skills, agents, and scaffolding in one place.
Learnings & gotchas
Collection of practical tips I've learnt as I go and gotchas to watch out for. See LEARNINGS_AND_GOTCHAS.md for the full list.
| Topic | Summary |
|---|---|
| Skills in cloud sessions | Project-scope your skills if you want them available in cloud-based autonomous sessions — user-level skills live on your machine and won't be cloned |
| Private repo access for Claude Code web | OAuth connection alone isn't enough — you must also grant access via https://github.com/apps/claude/installations/select_target |
| Agent SDK billing moves to plan-based (June 2025) | From 15 Jun 2025 you can use your Claude plan instead of a separate API key for cloud agents — migrate to get a consistent local/cloud auth pattern |
| No visual browser review in Claude Code web | Chrome/agent-browser is blocked in the web environment — do UI validation locally in the CLI or desktop app instead |
| Fable 5 can build non-trivial apps end-to-end in one session | Built a fully functional personal wiki from ~500 articles in ~10 mins |
Other helpful resources
| Resource | Description |
|---|---|
| Claude Code Mastery | Practical guide to getting the most out of Claude Code, including a curated list of popular CLAUDE.md files worth studying |
| Working with AI | Eugene Yan's reflections on day-to-day AI-assisted work — patterns, pitfalls, and how to stay in the driver's seat |
| autoresearch | Andrej Karpathy's autonomous research agent — a reference implementation for multi-step agentic research workflows |
| Levels of Agentic Engineering | A framework for thinking about how much autonomy to hand to AI agents at different stages of a project |
| Agentic Engineering Patterns | Simon Willison's guide to recurring patterns in agentic systems — practical and grounded in real-world experience |