programming

MySkills

🇧🇷 Versão em Português

Personal Claude Code skills marketplace, curated by Felipe Oliveira and organized into themed packages installable individually.

Why packages? Loading 200+ skills every session burns tokens for no reason. Here you install only the niche relevant to your current work.

Quick install

# Once per machine
/plugin marketplace add FelipeOFF/my-claude-code-skills

# Per work context — install only what you need now
/plugin install design@myskills
/plugin install copy@myskills
/plugin install marketing@myskills
/plugin install programming@myskills
/plugin install workflow@myskills

Packages

Package	Scope	What's inside	Details
design	UI/UX, design system, taste, mockups	Stitch (Google Labs), Huashu, frontend-design plugin, image generation, style packs, 21st.dev Magic MCP	→ design/README.md
copy	Copywriting, headlines, persuasive writing	scaffold (no curated skills yet)	→ copy/README.md
marketing	SEO, CRO, paid traffic, email, social	SEO audit (ECC), TopRank	→ marketing/README.md
programming	Backend, frontend web/mobile, testing, debugging, infra	ECC patterns (api-design, postgres, e2e), Stripe, Firebase, Agent SDK, Rust LSP, MCPs (chrome-devtools, react-grab, hostinger), TDD bundle	→ programming/README.md
workflow	Planning, memory, multi-phase, autonomous loops, multi-agent	superpowers, claude-mem, agentmemory, octo, codex, ralph-specum, sleepwell, GSD bundle (64 skills), find-skills, 1password	→ workflow/README.md

Why these skills?

This is active personal curation, not an exhaustive catalog. Every skill listed here was selected because the author actually uses it in real projects. Reasons cluster into:

• Override LLM defaults: design-taste skills, frontend-patterns, code-reviewer — LLMs default to generic outputs; these skills inject opinionated rules.
• Cross-session memory: claude-mem, agentmemory — long projects lose context without persistent memory.
• Multi-phase planning: GSD, ralph-specum, superpowers — features that span days need structure (PLAN.md, REVIEW.md, milestones).
• MCP integrations: chrome-devtools, hostinger, magic — tools the author uses daily get auto-configured (env vars stay personal).
• Stack patterns: Anthropic SDK, Postgres, REST design — opinionated patterns, not generic boilerplate.

Each package README explains why each included skill is in the curation.

Philosophy and rules

Every "which skill goes in which package" decision follows a written constitution: docs/CONSTITUTION.md.

Summary:

• Rule A — Single location. One skill, one package.
• Rule B — Top-level by default. Mid-level only with >15 skills + ≥5 cohesive cluster.
• Rule C — Frontmatter declares origin. authored | dependency | standalone-setup.
• Rule D — Standardized package README.
• Rule E — The Constitution is the source of truth.

Author's PT-BR/Jira workflow (repo root)

The directories agents/, commands/, and rules/ at the root are not part of any package. They're consumed via ~/.claude/CLAUDE.md global and support the author's personal workflow (Conventional Commits + Jira DTP-###, branches <type>/<JIRA>/<slug>, PRs in PT-BR).

• agents/ — commit-crafter, jira-linker, pr-writer
• commands/ — /branch, /commit, /pr, /review, /wt
• rules/ — branches, commits, context7, language, workflow

Contribute / curation

This is a personal curation. Issues with suggestions are welcome, but inclusion follows Constitution criteria. To open a PR adding a skill, read docs/CONSTITUTION.md first.