A skill = a muscle for a specific task.
feature-dev, smm, code-reviewer, frontend-design: what Claude Code skills are and which ones we actually use at AGINE every day.
Top skills for Claude
Skills are ready-made instructions for Claude Code. For example, the /smm skill runs trend research, writes a post, generates an image, and pushes it to 9 platforms in a single command. Set it up once and it works every day.
Grab our /smm skill
Here from the reel
The exact one the two of us use to run 12 social accounts. A Markdown file, ready to go. Download it, drop it into your project, run it.
⬇ Download smm.md · How to install in 3 steps ↓
Download not working? Open the link: guides.agineai.com/skills/smm.md and save it via File → Save As.
Installing /smm in 3 steps
If you're in Claude Code (CLI)
01. Download smm.md using the button above It lands in your default Downloads folder.
02. Drop it into your project
Create a .claude/skills/ folder in your project root (if you don't have one yet) and put smm.md there.
``
mkdir -p .claude/skills && mv ~/Downloads/smm.md .claude/skills/
``
03. Launch Claude Code and call the command In your terminal:
``
claude
> /smm
``
The skill will ask for API keys (ScrapeCreators, Telegram) on its own if they're missing from .env, then run the whole workflow.
No Claude Code CLI? It installs in 5 minutes: see the Claude Code guide. Or take the simpler route: open the contents of smm.md and paste them into Claude Projects as a System Prompt. You'll get similar behavior without the CLI.
Who this guide is for Skills work in Claude Code, the terminal tool for developers and power users. If you're just getting started, begin with setting up Claude in 10 steps or how the Content Factory works.
What a skill is
Describe it once and it works the same way every time.
A skill is a markdown file with instructions for Claude. It lives in .claude/skills/smm.md. When you call /smm, Claude reads that file and executes the workflow described inside.
Skills can launch subagents: parallel helpers for different parts of the task. feature-dev runs code-architect + code-explorer + code-reviewer at the same time.
``` # .claude/skills/ structure
.claude/ └── skills/ ├── smm.md ← called as /smm ├── feature-dev.md ← called as /feature-dev ├── code-reviewer.md ← called as /review └── hookify.md ← called as /hookify ```
Our working library
The skills we use every day.
/smm: SMM Content Factory _Content for 9 platforms in 2 hours a day_
The full content-creation workflow: trend research via ScrapeCreators, reference deconstruction, safe remix mechanics, AI art direction, image generation, publishing through n8n. All in one command.
Runs these agents: trend-researcher → content-writer → art-director → publisher
When to use it: every day when we're making content. This is our main production skill.
⬇ Download the .md (public version)
/feature-dev: Feature Development _Features end to end, with tests and review_
Orchestrates three specialized agents: code-architect designs the architecture, code-explorer digs through the existing code, code-reviewer checks the result. Each one works in its own context, in parallel.
Runs these agents: code-architect + code-explorer (in parallel) → code-reviewer
When to use it: a new feature, a module refactor, any task that needs architectural thinking before the code gets written.
/review: Code Reviewer _PR review to a production-quality standard_
Systematic code review with a confidence filter: it only surfaces high-confidence findings. Checks bugs, security, style, tests. It doesn't flood you with the obvious, just what actually matters.
When to use it: before every commit, after writing a new component, before opening a PR.
/hookify: Hookify _Creates hooks from your conversation_
Analyzes the current conversation, finds the Claude behaviors that annoyed you, and suggests hooks to prevent them in the future. It learns from your frustrations.
When to use it: when Claude did something wrong and you want to make sure it never happens again.
/frontend-design: Frontend Design _Production-grade UI components_
Builds distinctive interfaces with high design quality. Not just working code: components with proper typography, proportions, and hover states. The kind of thing you can show a client right away.
When to use it: a new UI component, a landing page, a dashboard. Anywhere you need a design result, not just markup.
/simplify: Simplify _Refactoring without changing functionality_
Goes through recently changed code, removes duplication, simplifies logic, improves readability, and leaves alone what shouldn't be touched. It focuses on the changed files, not the whole project.
When to use it: after writing new code, before a PR, whenever you need to bring code up to standard without risking breakage.
/commit-push-pr: Commit + Push + PR _Three steps in one command_
Looks at your changes, writes a meaningful commit message, runs git push, and opens a PR with a description. No need to think about wording: Claude writes up what changed and why.
When to use it: when a feature is done and you want it packaged into a PR without burning time on manual git flow.
How to install a skill
Three steps. Works right away.
01. Create the folder
If it's not there yet, create .claude/skills/ in your project root.
02. Drop in the .md file
Copy a ready-made skill from GitHub or write your own. File name = command name. The file smm.md is called as /smm.
03. Call it in a session
Type /skill-name in Claude Code and it will execute everything described in the .md. No restart needed.