Skills - Coleção de Skills para Agentes IA
vtmocanu/skills
Skills modulares para ampliar capacidades de agentes Claude e dot-ai
Ver no GitHubRepositório público com uma coleção de skills reutilizáveis para agentes do dot-ai e Claude Code. Inclui skills para automação de terminal (cmux), gestão de permissões de agentes, criação de times de agentes, navegação e customização. As skills são versionadas via YAML frontmatter e podem ser injetadas em agentes através da CLI dot-ai.
Instalação
npx convez-ai vtmocanu/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
- Estender funcionalidades de agentes com skills de automação terminal e browser
- Orquestrar times de agentes para tarefas complexas em repositórios
- Personalizar ambiente cmux (multiplexador de terminal) via skills
Veredito

Coleção técnica bem estruturada de skills reutilizáveis para o ecossistema dot-ai/Claude Code, focada em automação terminal, orquestração de agentes e customização de ambientes. Adequada para desenvolvedores que querem estender agentes com capacidades específicas de infraestrutura e automação.
README
A tradução para o português deste README ainda está na fila.
skills
A public collection of agent skills for dot-ai and Claude Code.
Each skill lives at the repository root, either as a single Markdown file (<name>.md) or as a folder (<name>/SKILL.md plus supporting files; supported since dot-ai v1.21.0), with YAML frontmatter (name + description). dot-ai serves them to your agent: it fetches this repo and generates the skill files into your agent's skills directory. (Skills live at the root because dot-ai's ?repo= override reads prompts from the repository root.)
Quick Start
Generate these skills into Claude Code with the dot-ai CLI (v1.21.0+), pointed at a running dot-ai server:
dot-ai skills generate --agent claude-code --repo https://github.com/vtmocanu/skills
--repo composes alongside other sources: each invocation tags its skills with source: frontmatter and rewrites only its own slice, so running it once per repo (typically one agent hook per source) lets skills from several repos coexist without clobbering each other.
Skills
| Skill | What it does |
|---|---|
| agent-permissions | Manage an AI coding agent's permissions via Dippy (Bash/MCP allow/ask/deny + the auto-mode [ASK] fallback wrapper, bundled) and settings.json (Read/WebFetch/Skill). |
| agent-team | Auto-generate and run a per-repo Claude Code agent team: probe the repo, write .claude/agents/{role}.md subagent definitions from a role library, then orchestrate tasks with TeamCreate plus spawned teammates. |
| cmux | Control cmux topology and routing: windows, workspaces, panes/surfaces, focus, moves, reorder, identify, trigger flash. |
| cmux-browser | Browser automation with cmux: open sites, interact with pages, wait for state changes, and extract data from cmux browser surfaces. |
| cmux-customization | Customize cmux: cmux.json actions, custom commands, workspace layouts, buttons, Command Palette, shortcuts, notifications, browser routing, and presets. |
| cmux-diagnostics | Run cmux diagnostics and health checks when hooks, notifications, session restore, settings, socket access, or CLI control are not working. |
| cmux-keyboard-shortcuts | Customize, rebind, unbind, reset, audit, or template cmux keyboard shortcuts (tmux-style, Vim-style, terminal-first, and more). |
| cmux-markdown | Open markdown files in a formatted viewer panel with live reload, alongside the terminal. |
| cmux-settings | View and edit cmux settings in ~/.config/cmux/cmux.json: set values by JSON path, validate the file, and look up recognized keys. |
| cmux-workspace | Work inside the current cmux workspace and terminal: caller surface, panes, surfaces, socket targeting, and non-interfering automation. |
| reflect | Analyze the current session and propose improvements to the skill that was used, then edit and commit it. |
Other files
CLAUDE.example.md— a generic starter for~/.claude/CLAUDE.md(Claude Code's global instructions): general AI-collaboration guidance only, no setup specifics. Not a skill; copy what's useful into your own config.
Contributing
Issues and PRs welcome. See CONTRIBUTING, the Code of Conduct, and the Security Policy.
License
MIT © Vlad Mocanu