Skills para Agentes de Código
hoangtmbk/skills
Slash commands compostos para agentes de código — alinhamento, QA, e engenharia real
Ver no GitHubConjunto de skills (slash commands) reutilizáveis para Claude Code, Codex e outros agentes de IA, focando em alinhamento, qualidade e engenharia prática. Inclui ferramentas para validação com browser, perguntas exploratórias (/grill-me) e integração com documentação de projeto.
Instalação
npx convez-ai hoangtmbk/skills1 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
- Alinhar requisitos antes de gerar código com /grill-me e /grill-with-docs
- Validar comportamento de aplicações via browser QA com Playwright
- Estruturar fluxos de engenharia reproducíveis em projetos
- Integrar agentes de IA em pipelines de desenvolvimento com suporte a múltiplos modelos
Veredito

Skills sólidas e bem pensadas para reduzir o ruído entre intenção e execução em desenvolvimento com IA. A abordagem de 'grilling' (exploração de requisitos) é uma das ideias mais práticas do projeto. Recomendado para equipes que usam Claude Code regularmente e buscam workflow reprodutível.
README
A tradução para o português deste README ainda está na fila.
Skills
Agent skills for real engineering — not vibe coding. Daily-driver slash commands for Claude Code, Codex, and other coding agents. Forked from mattpocock/skills and adapted for my workflow.
Developing real applications is hard. Approaches like GSD, BMAD, and Spec-Kit try to help by owning the process. But while doing so, they take away your control and make bugs in the process hard to resolve.
These skills are designed to be small, easy to adapt, and composable. They work with any model. They're based on decades of engineering experience. Hack around with them. Make them your own.
Quickstart
Claude Code (recommended — plugin install)
Ships skills + the bowser-qa-agent subagent + the /browser-qa and /qa-clean slash commands as one Claude Code plugin.
One-time per machine — register the marketplace:
/plugin marketplace add hoangtmbk/skills
Per target repo — install the plugin (project scope):
/plugin install hoangta-skills@hoangtmbk-skills
Restart Claude Code after install. Then run /setup-hoangta-skills once per repo to wire up docs/agents/ config (issue tracker, triage label vocabulary, doc location).
Codex, Gemini CLI, other coding agents (skills only)
The npx skills@latest installer drops the skills/ directory into .agents/skills/. It does not install the bowser-qa-agent subagent or the /browser-qa / /qa-clean slash commands — those are Claude Code plugin-only. If you need the browser-QA harness on a non-Claude-Code agent, copy the files manually per skills/engineering/playwright-qa/INSTALL.md.
npx skills@latest add hoangtmbk/skills
[!NOTE] Using the npx installer on Claude Code (fallback only): Claude Code auto-discovers skills under
.claude/skills/, not.agents/skills/. After install, symlink them across:mkdir -p .claude/skills && \ for d in .agents/skills/*/; do ln -sfn "$PWD/${d%/}" ".claude/skills/$(basename "$d")"; doneThe plugin install above is the supported path for Claude Code — use this only if the plugin install isn't an option.
Then run /setup-hoangta-skills to set up docs/agents/ config.
Why These Skills Exist
These skills fix common failure modes seen with Claude Code, Codex, and other coding agents.
#1: The Agent Didn't Do What I Want
"No-one knows exactly what they want"
David Thomas & Andrew Hunt, The Pragmatic Programmer
The Problem. The most common failure mode in software development is misalignment. You think the dev knows what you want. Then you see what they've built - and you realize it didn't understand you at all.
This is just the same in the AI age. There is a communication gap between you and the agent. The fix for this is a grilling session - getting the agent to ask you detailed questions about what you're building.
The Fix is to use:
/grill-me- for non-code uses/grill-with-docs- same as/grill-me, but adds more goodies (see below)
These are the most popular skills in the set. They help you align with the agent before getting started, and think deeply about the change you're making. Use them every time you want to make a change.
#2: The Agent Is Way Too Verbose
With a ubiquitous language, conversations among developers and expressions of the code are all derived from the same domain model.
Eric Evans, Domain-Driven-Design
The Problem: At the start of a project, devs and the people they're building the software for (the domain experts) are usually speaking different languages.
The same tension shows up with agents. Agents are usually dropped into a project and asked to figure out the jargon as they go. So they use 20 words where 1 will do.
The Fix for this is a shared language. It's a document that helps agents decode the jargon used in the project.
Here's an example CONTEXT.md, from Matt Pocock's course-video-manager repo. Which one is easier to read?
- BEFORE: "There's a problem when a lesson inside a section of a course is made 'real' (i.e. given a spot in the file system)"
- AFTER: "There's a problem with the materialization cascade"
This concision pays off session after session.
This is built into /grill-with-docs. It's a grilling session that helps you build a shared language with the AI, and document hard-to-explain decisions in ADR's.
It's hard to explain how powerful this is. It might be the single coolest technique in this repo. Try it, and see.
[!TIP] A shared language has many other benefits than reducing verbosity:
- Variables, functions and files are named consistently, using the shared language
- As a result, the codebase is easier to navigate for the agent
- The agent also spends fewer tokens on thinking, because it has access to a more concise language
#3: The Code Doesn't Work
"Always take small, deliberate steps. The rate of feedback is your speed limit. Never take on a task that's too big."
David Thomas & Andrew Hunt, The Pragmatic Programmer
The Problem: Let's say that you and the agent are aligned on what to build. What happens when the agent still produces crap?
It's time to look at your feedback loops. Without feedback on how the code it produces actually runs, the agent will be flying blind.
The Fix: You need the usual tranche of feedback loops: static types, browser access, and automated tests.
For automated tests, a red-green-refactor loop is critical. This is where the agent writes a failing test first, then fixes the test. This helps give the agent a consistent level of feedback that results in far better code.
The /tdd skill can slot into any project. It encourages red-green-refactor and gives the agent plenty of guidance on what makes good and bad tests.
For debugging, the /diagnose skill wraps best debugging practices into a simple loop.
#4: We Built A Ball Of Mud
"Invest in the design of the system every day."
Kent Beck, Extreme Programming Explained
"The best modules are deep. They allow a lot of functionality to be accessed through a simple interface."
John Ousterhout, A Philosophy Of Software Design
The Problem: Most apps built with agents are complex and hard to change. Because agents can radically speed up coding, they also accelerate software entropy. Codebases get more complex at an unprecedented rate.
The Fix for this is a radical new approach to AI-powered development: caring about the design of the code.
This is built in to every layer of these skills:
/to-prdquizzes you about which modules you're touching before creating a PRD/zoom-outtells the agent to explain code in the context of the whole system
And crucially, /improve-codebase-architecture helps you rescue a codebase that has become a ball of mud. Run it on your codebase once every few days.
Summary
Software engineering fundamentals matter more than ever. These skills are a condensation of those fundamentals into repeatable practices, to help you ship the best apps of your career.
Reference
Engineering
Skills I use daily for code work.
- diagnose — Disciplined diagnosis loop for hard bugs and performance regressions: reproduce → minimise → hypothesise → instrument → fix → regression-test.
- grill-with-docs — Grilling session that challenges your plan against the existing domain model, sharpens terminology, and updates
CONTEXT.mdand ADRs inline. - triage — Triage issues through a state machine of triage roles.
- improve-codebase-architecture — Find deepening opportunities in a codebase, informed by the domain language in
CONTEXT.mdand the decisions indocs/adr/. - playwright-qa — Headless functional QA of running web apps. Pairs with the
bowser-qa-agentsubagent and the/browser-qa+/qa-cleanslash commands (all ship in this plugin) to fan out YAML user-stories in parallel and produce PASS/FAIL Run reports with PNG, console, network, and trace artifacts. - setup-hoangta-skills — Scaffold the per-repo config (issue tracker, triage label vocabulary, domain doc layout) that the other engineering skills consume. Run once per repo before using
to-issues,to-prd,triage,diagnose,tdd,improve-codebase-architecture, orzoom-out. - tdd — Test-driven development with a red-green-refactor loop. Builds features or fixes bugs one vertical slice at a time.
- to-issues — Break any plan, spec, or PRD into independently-grabbable GitHub issues using vertical slices.
- to-prd — Turn the current conversation context into a PRD and submit it as a GitHub issue. No interview — just synthesizes what you've already discussed.
- zoom-out — Tell the agent to zoom out and give broader context or a higher-level perspective on an unfamiliar section of code.
- prototype — Build a throwaway prototype to flesh out a design — either a runnable terminal app for state/business-logic questions, or several radically different UI variations toggleable from one route.
Productivity
General workflow tools, not code-specific.
- caveman — Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler while keeping full technical accuracy.
- grill-me — Get relentlessly interviewed about a plan or design until every branch of the decision tree is resolved.
- handoff — Compact the current conversation into a handoff document so another agent can continue the work.
- teach — Teach the user a new skill or concept over multiple sessions, using the current directory as a stateful teaching workspace.
- write-a-skill — Create new skills with proper structure, progressive disclosure, and bundled resources.
Misc
Tools I keep around but rarely use.
- git-guardrails-claude-code — Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, etc.) before they execute.
- migrate-to-shoehorn — Migrate test files from
astype assertions to @total-typescript/shoehorn. - scaffold-exercises — Create exercise directory structures with sections, problems, solutions, and explainers.
- setup-pre-commit — Set up Husky pre-commit hooks with lint-staged, Prettier, type checking, and tests.
Attribution
Based on mattpocock/skills by Matt Pocock. The skill content, the "Why These Skills Exist" essay, and the overall structural design are his work. This fork strips upstream-specific branding and adapts the setup skill for my own workflow. MIT licensed — see LICENSE.