Telusko Docs

Understand skills, the feature that turns repeated prompts into your own slash commands, how they relate to custom commands, and when to use one.

Type / in Claude Code and you see commands like /init, /clear and /compact. You have also seen bundled skills like /code-review. Now we create our own.

The problem: the same long prompt, again and again

Remember the prompt we used to build the CRUD endpoints?

Read @specs/api-contract.md.
Build the four FastAPI endpoints in backend/app/api/v1/tasks.py ...
Business logic goes in backend/app/services/task_service.py.
Routes know nothing about SQLAlchemy. Service knows nothing about HTTP.
Response shapes must match exactly what is in the spec.
Handle errors with HTTPException using the status codes from the spec.
...

It works well because it carries all our rules. But the project will keep growing: a login endpoint, filtering, pagination. Each time, we would need to type (or paste) the same rules again, and hope we did not forget a line.

What if we could just type:

/scaffold-endpoint GET /api/tasks/by-role/{role}

and have all those rules applied automatically? That is exactly what a skill gives you.

What is a skill?

A skill is a folder with a SKILL.md file inside. The file has two parts:

YAML frontmatter: metadata such as a description and which tools it may use
Markdown instructions: what Claude should do when the skill runs

The folder name becomes the command. A skill at .claude/skills/scaffold-endpoint/SKILL.md becomes /scaffold-endpoint.

.claude/
└── skills/
    ├── telusko-status/
    │   └── SKILL.md
    ├── scaffold-endpoint/
    │   └── SKILL.md
    └── implement-spec/
        └── SKILL.md

Custom commands vs skills

Earlier versions of Claude Code had two separate features: custom commands (a markdown file in .claude/commands/) and skills. They have now been merged:

	`.claude/commands/deploy.md`	`.claude/skills/deploy/SKILL.md`
Creates the command	`/deploy`	`/deploy`
Still supported	Yes	Yes, and recommended
Can include extra files (scripts, templates)	No	Yes, inside the skill folder

So when you want a custom command today, you create a skill.

Two ways a skill gets used

Who triggers it	How
You	Type `/skill-name`, optionally with arguments
Claude	Claude reads every skill's `description` at the start of a session. When your request matches, it loads the skill on its own

If a skill should only run when you ask for it (for example, anything that changes code or deploys), add disable-model-invocation: true to its frontmatter.

Skills are cheap on context

Unlike CLAUDE.md, which loads in full every session, only the short description of each skill is loaded upfront. The full instructions load only when the skill is used. So you can have many skills without filling your context window.

	`CLAUDE.md`	Skill
Loads	Every session, in full	Description always, body only when used
Best for	Facts Claude should always know	Procedures, checklists, repeatable workflows
Example	"Use uv, not pip"	"How to add a new endpoint"

A good rule from the official docs: when a section of CLAUDE.md turns from a fact into a procedure, move it into a skill.

Bundled skills worth knowing

Claude Code ships with prompt-based skills you can use right away:

Skill	What it does
`/run`	Launches your app so you can see a change working
`/verify`	Builds and runs your app to confirm a change does what it should
`/run-skill-generator`	Records how to start your project as a project skill, so `/run` follows it every time
`/code-review`	Reviews the current diff for bugs
`/debug`	Turns on debug logging and helps troubleshoot Claude Code itself
`/simplify`	Reviews recent changes for cleanup opportunities and applies fixes

A skill for running the project

In the video, we repeatedly typed "run the frontend and the backend". /run and /run-skill-generator solve exactly that: record once how the project starts (environment variables, ports, order), and every later run follows the same steps. It is like the "How to run" section of a README, except Claude actually executes it.