davila7/claude-with-skills

Perfil de davila7

Última atividade de origem registrada: 15 de maio de 2026 às 01:57

Catálogo do SkillsMP atualizado: 17 de outubro de 2026 às 00:09

skills coletadas: 50

Estrelas no GitHub: 13
Forks no GitHub: 0

README do repositório

Arquivo de origem do README: README.md

Ler o README original

Expandir README completo

# claude-with-skills <img width="1055" height="564" alt="Screenshot 2026-05-15 at 09 53 22" src="https://github.com/user-attachments/assets/4c0ba288-e1c5-4299-b5f6-57a3c9c982b4" /> A progressive course for developers who want to stop pasting the same instructions into Claude and start shipping reusable, portable Agent Skills. Whether you are tired of re-explaining your release process every Monday, migrating existing slash commands to the new Skills format, building a plugin for your team, or just learning what Claude Code can do — this repo takes you from your first `SKILL.md` to advanced patterns like forked context, dynamic injection, and headless automation. ## Quickstart ```bash # 1. Clone the repo git clone https://github.com/davila7/claude-with-skills.git cd claude-with-skills # 2. Install the hello-skill example into your personal skills directory bash scripts/install-examples.sh src/content/docs/01-basic/lesson-01-anatomy/examples/hello-skill personal # 3. Open Claude Code and invoke the skill claude # then type: /hello-skill ``` That is it. You have just installed and invoked your first Agent Skill. ## Learning path | Level | Section | Outcome | |---|---|---| | Introduction | [00-introduction/](src/content/docs/00-introduction/index.md) | Understand what skills are, how they load, and how they compare to CLAUDE.md, subagents, and MCP | | Basic | [01-basic/](src/content/docs/01-basic/index.md) | Write a skill from scratch, add frontmatter, invoke it by name and with arguments | | Intermediate | [02-intermediate/](src/content/docs/02-intermediate/index.md) | Use supporting files, dynamic context injection, forked context, and hooks | | Advanced | [03-advanced/](src/content/docs/03-advanced/index.md) | Build agent-driven skills, headless pipelines, plugin packaging, and cross-tool portability | You can also browse the course as a **website** built with [Astro Starlight](https://starlight.astro.build): ```bash npm install npm run dev # then open http://localhost:4321/ ``` ## How to run examples ### Interactive mode ```bash claude ``` Inside the interactive session, invoke any installed skill by name: ``` /skill-name /skill-name argument ``` ### Headless mode Pass a prompt directly with `-p`. Claude processes it and exits. ```bash # Invoke a skill by name claude -p "/skill-name" # Invoke a skill with an argument claude -p "/skill-name argument" # Pipe context into a skill echo "content to process" | claude -p "/skill-name" ``` Headless mode is useful in scripts, CI pipelines, and Makefile targets where you want Claude to perform a task without an interactive session. Output goes to stdout. ## AgentSkills open standard vs Claude Code Skills follow the [AgentSkills.io open standard](https://agentskills.io). The standard defines three portable fields that every compliant tool understands: ```yaml --- name: your-skill-name description: What the skill does and when to invoke it. allowed-tools: - Bash - Read --- ``` If you write only these fields, your skill runs unchanged in Cursor, GitHub Copilot, Gemini CLI, and any other AgentSkills-compatible tool. Claude Code implements the full standard and adds a superset of extra fields for finer control: | Field | What it does | |---|---| | `disable-model-invocation` | Prevent automatic activation; require explicit invocation | | `user-invocable` | Control whether users can invoke the skill by typing `/skill-name` | | `argument-hint` | Hint shown to the user when the skill expects an argument | | `arguments` | Declare named, typed arguments with descriptions | | `paths` | Restrict the skill to specific file paths or patterns | | `shell` | Run a shell command and inject the output into context | | `context: fork` | Execute the skill in an isolated context window | | `agent` | Configure subagent behavior | | `hooks` | Attach lifecycle hooks (before, after, on-error) | | `model` | Specify which Claude model to use for this skill | | `effort` | Control reasoning effort (low / medium / high) | Dynamic context injection with `` !`command` `` and `${CLAUDE_SKILL_DIR}` are also Claude Code-specific. These extra fields are silently ignored by tools that do not support them, so your skill degrades gracefully in other environments. ## Reference - [Frontmatter cheatsheet](src/content/docs/reference/frontmatter-cheatsheet.md) — every field, its type, and an example value - [Headless mode guide](src/content/docs/reference/headless-mode.md) — flags, patterns, piping, and CI integration - [Troubleshooting](src/content/docs/reference/troubleshooting.md) — skill not found, invocation not working, context issues ## Contributing See [CONTRIBUTING.md](CONTRIBUTING.md) for naming conventions, frontmatter rules, and PR expectations. ## License MIT License. See [LICENSE](LICENSE).

Carregando o README do repositório…

Skills neste repositório

Desenvolvedores de software Administradores de redes e sistemas de computador Analistas de garantia de qualidade de software e testadores Analistas de segurança da informação

9 categorias ocupacionais · 100% classificado

skill

ocupação

descrição

atualizado

explain-code

Desenvolvedores de software

Explains code in plain language for someone unfamiliar with the programming language. Use when asked to explain code, walk through logic, describe what a function does, or when the user says "explain this" or "walk me through this".