Stranger Skill
Jeyanarendran/stranger-skill
Ciclo automatizado de build, review e test para código robusto
Ver no GitHubUma skill para Claude Code que automatiza o ciclo build → review → test usado em produção para enviar código sem regressões. Encapsula práticas reais de um desenvolvedor solo que mantém aplicações multi-surface, com regras concretas derivadas de falhas reais em produção.
Instalação
npx convez-ai Jeyanarendran/stranger-skill1 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 validação de código antes de merge
- Implementar testes sistemáticos em pipelines CI/CD
- Revisar próprio código com checklist estruturado
Veredito

Skill pragmática que traduz experiência real em automação testável. Útil para equipes que querem rigor operacional sem overhead, mas demanda compreensão do modelo mental build-review-test para aproveitar plenamente.
README
A tradução para o português deste README ainda está na fila.
stranger-skill
A Claude Code skill that encodes the build → review → test loop I actually use day-to-day — distilled from shipping DotMD (a Markdown-native productivity suite: docs, slides, sheets, native shells, AI everywhere). It is opinionated, repeatable, and biased toward shipping working code without regressions.
What's inside
skills/stranger-skill/SKILL.md— the skill itself, organized as three phases (Build, Review, Test) with concrete steps and the rules that prevent the failure modes I've actually hit.
Installing
Drop the skills/stranger-skill/ directory into your Claude Code skills path (~/.claude/skills/ or the project-local .claude/skills/). Invoke from a chat with /stranger-skill or let Claude auto-invoke when the task description matches.
Why this exists
Most "best practice" guides describe what should happen. This one captures what actually happens when a single dev ships a multi-surface app, reviews their own PRs, and runs the full test pipeline. Every rule here exists because it prevented (or was learned from) a real outage.
License
MIT
Used by DotMD for live GitHub testing
This repo is intentionally minimal and public so it can serve as a stable target for a fine-grained PAT in DotMD's e2e + manual test flows (Settings -> GitHub -> live repo + branch dropdowns -> Sync).
If you're a DotMD agent, see AGENTS.md -> GitHub Settings test PAT in the marc-prod monorepo for the create recipe. The PAT lives ONLY in dotmd/apps/web/.env.local as GITHUB_PAT_STRANGER_SKILL=... (gitignored). It is scoped to this repo only (Contents R/W, Metadata R) so the blast radius of a leak is contained to a public docs repo with no secrets.