8 Skills That Changed How I Ship
A curated collection of Claude Code Skills for indie developers who want to build faster, test smarter, and stop doing everything by hand.
This is a companion piece to my series on building a professional AI coding stack (Skills, Not Prompts, AGENTS.md). Those articles covered the architecture. This one covers the gear.
The Skills ecosystem for Claude Code went from zero to over 60,000 published skills in five months. That’s exciting. It’s also overwhelming.
Most developers I talk to installed one or two skills, maybe the official document suite, and stopped there. Not because the ecosystem isn’t useful - but because nobody curated it for them. Nobody said “these eight, in this order, for this reason.”
That’s what this post is.
I’ve been testing skills obsessively across my own projects - Kelo (subscription tracker), Hoople (social network), and client work at Juice Studio. What follows is the collection that stuck. The ones that changed how I work day to day. Not a directory dump. Not a “top 50 awesome list.” Eight skills across four categories, with honest notes on what each one actually does and why it earned a permanent spot in my setup.
Ship the Product
1 - Frontend Design
Anthropic’s official answer to every AI-built interface looking exactly the same.
Everything Claude Code builds tends to converge on the same aesthetic: Inter font, purple gradients, predictable card layouts, safe grays. Anthropic calls this “AI slop” - distributional convergence toward the statistical center of design decisions. This skill exists to break that pattern.
Before writing a single line of code, the skill forces Claude to commit to a bold aesthetic direction across four dimensions: purpose, tone, constraints, and differentiation. It bans generic fonts (Inter, Roboto, Arial, Space Grotesk), demands cohesive color systems with dominant colors and sharp accents, and pushes for meaningful motion - orchestrated page reveals, not scattered micro-interactions.
The skill includes 50 design styles, curated typography pairings, and an explicit anti-slop philosophy. One developer ran a blind evaluation across 50 prompts: the skill won 8 out of 10 comparisons on creative distinction alone.
You don’t have a design team. Your app’s first impression is its only impression. This skill is the difference between “looks like every other AI-built landing page” and “wait, who designed this?”
Source: anthropics/claude-code (official) | Installs: 277K+
npx skills add anthropics/claude-code --skill frontend-design2 - App Store Connect CLI
From zero configuration to TestFlight public beta in under two minutes. No browser required.
If you ship iOS apps, you know the pain. App Store Connect is a slow, click-heavy web interface that makes every release feel like paperwork. This CLI - written in Go, JSON-first output, designed explicitly for AI agents - replaces all of it.
TestFlight uploads, build management, metadata sync, signing, certificates, provisioning profiles, screenshot pipelines, submission health checks, and release staging. All from your terminal. All scriptable. All agent-friendly.
Thomas Ricouard (Ice Cubes, Codex Monitor) documented going from nothing configured to a TestFlight public beta in under two minutes using this tool. His exact words: “The agent literally did everything itself.”
The screenshot pipeline alone is worth the install. Define a JSON plan, the agent captures screens from the simulator, frames them, and prepares them for upload. What used to take an afternoon now takes minutes.
You’re probably spending 2-3 hours per release on App Store Connect tasks. This gives those hours back.
Source: rudrankriyam/App-Store-Connect-CLI | GitHub Stars: 728 | Agent Skills: 15+
brew install rudrankriyam/tap/asc
npx skills add rudrankriyam/app-store-connect-cli-skills3 - Remotion
Programmatic video creation with React. Describe what you want, Claude writes the code, you get an MP4.
This isn’t AI generating random video from text prompts. Remotion turns video into a React project - each frame is a component, each animation is an interpolation function, and the output renders to MP4. With this skill installed, Claude Code becomes your video editor.
Describe what you want in plain English - “a 10-second title animation with spring physics and a gradient background” - and Claude generates the TypeScript, the composition config, the timing, and the render pipeline. The skill covers 20+ rule categories: 3D content, animations, captions, audio handling, chart visualizations, light leaks, and more.
One early adopter documented building a complete product video in 2 hours through conversational refinement with Claude. The launch demo alone hit 6M+ views.
Short-form video is the distribution channel of 2026. If you can’t show your app’s main benefit in a 4-second screen recording, you’re invisible. Remotion + Claude Code makes that trivially fast.
Source: remotion-dev/skills | Weekly installs: 126K+ | Launch demo views: 6M+
npx skills add remotion-dev/skillsWrite Bulletproof Code
4 - TDD (Test Driven Development)
If you write code before tests, this skill deletes it. Not a metaphor.
Claude’s natural instinct is to write implementation first, then add tests as validation. This defeats the entire purpose of TDD - tests should drive design, not confirm existing code. The TDD skill from Superpowers inverts that instinct by making it structurally impossible to skip the Red-Green-Refactor cycle.
Write a failing test first (Red). Write the minimum code to make it pass (Green). Clean up while tests stay green (Refactor). Every single time. No shortcuts. If you try to skip the failing test step, the agent refuses to proceed.
The result: 100% regression coverage on new features. Every function starts with a specification of what it should do before a single line of implementation exists. Bug fixes start with a test that reproduces the bug. Refactors are safe because the test suite catches regressions instantly.
You don’t have QA. You don’t have code reviewers. TDD is your safety net - and this skill makes the safety net automatic.
Source: obra/superpowers | GitHub Stars: 99K+ (Anthropic marketplace verified)
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace5 - Systematic Debugging
No more random code changes hoping something sticks. A four-phase scientific method for every bug.
Without this skill, Claude tends to patch symptoms rather than fix causes. It sees an error, changes the line that produced it, and moves on. The debugging skill from Superpowers forces a structured approach:
Phase 1: Form a hypothesis about the root cause.
Phase 2: Gather evidence - logs, stack traces, reproduction steps.
Phase 3: Test the hypothesis, verify the fix.
Phase 4: Document the root cause and prevent recurrence.
The skill catches root causes that ad-hoc debugging misses, reduces revision rounds by roughly 30%, and works particularly well on complex bugs and intermittent failures - the kind that make you question your career choices at 2 AM.
Every hour spent debugging is an hour not spent shipping. A structured approach is the fastest way out of the hole and back to building.
Source: obra/superpowers (included in TDD install above)
6 - Webapp Testing
Playwright-powered E2E testing. Claude navigates your app, clicks flows, fills forms, takes screenshots, and reports what broke.
Describe what you want tested - “verify the signup flow from landing page to dashboard” - and Claude generates the Playwright script, runs it against your local dev server, captures screenshots at each step, and reports what passed and what broke.
Two primary use cases: E2E verification (does the happy path work?) and QA regression (did my last change break something visually?). Both are critical. Both are tedious to do manually. Both are exactly the kind of deterministic, repeatable work that agents excel at.
Manual QA doesn’t scale when you’re shipping alone. This skill turns “I think the signup flow still works” into “I verified it works, and here are the screenshots to prove it.”
Source: anthropics/claude-code (official)
npx skills add anthropics/claude-code --skill webapp-testingThink Before You Build
7 - Brainstorming
The most underrated skill in the entire collection. Forces structured exploration before any code gets written.
Instead of jumping to implementation, Claude initiates a Socratic dialogue: identify constraints, list edge cases and alternative approaches, evaluate trade-offs before you commit to an architecture. The skill asks questions, challenges assumptions, explores technological alternatives, and only produces a design recommendation after it understands the problem space.
Features built after brainstorming produce roughly 30% fewer revision rounds. That’s the difference between building the right thing once and rebuilding the wrong thing three times.
The skill activates automatically when you describe a new feature or project. Ten minutes of structured thinking saves hours of wasted implementation.
Your biggest bottleneck isn’t coding speed - it’s building the wrong thing. This is the cheapest insurance against wasted effort.
Source: obra/superpowers (included in TDD install above)
Market What You Build
8 - Marketing Skills
32 skills. 52 CLI tools. The most comprehensive marketing collection in the ecosystem, built for founders who ship alone.
Built by Corey Haines (Conversion Factory) for technical founders who need marketing help but don’t have a marketing team. Seven categories: CRO, copywriting, SEO, analytics, email sequences, content strategy, and growth engineering.
The skills work with a shared product marketing context file (.agents/product-marketing-context.md) that you fill in once with your product, audience, positioning, and competitive landscape. Every skill reads that context first, so you never repeat yourself.
Practical examples: “Help me optimize this landing page for conversions” triggers page-cro. “Write homepage copy” triggers copywriting. “Set up GA4 tracking for signups” triggers analytics-tracking. “Create a 5-email welcome sequence” triggers email-sequence. Each skill includes frameworks, best practices, and the kind of specific guidance that generic prompting can’t match.
The tool registry includes integration guides for GA4, Stripe, Mailchimp, Google Ads, Resend, Zapier, and more - all agent-compatible. Works with Claude Code, Codex, Cursor, and any agent supporting the Agent Skills spec.
You can build the best product in the world and still fail if nobody finds it. These skills turn Claude into a marketing co-pilot that actually knows what it’s doing.
Source: coreyhaines31/marketingskills | GitHub Stars: 17.8K+ | Skills: 32 | CLI Tools: 52
npx skills add coreyhaines31/marketingskillsHow I Organize This
I don’t install all eight skills globally. I organize by project context:
For iOS apps (Kelo, Hoople): Frontend Design + App Store Connect CLI + TDD + Systematic Debugging. Webapp Testing joins when there’s a web component.
For client work (Juice Studio): Frontend Design (every project) + Brainstorming (every new feature) + Marketing Skills (when the deliverable includes growth work).
For content creation: Remotion (video for social media) + Marketing Skills (distribution strategy).
The key insight from my Skills article: skills load on-demand, not always-present. They don’t consume instruction budget until they activate. Having 15 skills installed but only using 3 per session costs you almost nothing.
The Install Cheatsheet
Everything in one block. Under 10 minutes to set up the complete stack:
bash
# Official Anthropic skills
npx skills add anthropics/claude-code --skill frontend-design
npx skills add anthropics/claude-code --skill webapp-testing
# Superpowers (TDD + Debugging + Brainstorming + 20 more)
# Run inside Claude Code:
/plugin marketplace add obra/superpowers-marketplace
/plugin install superpowers@superpowers-marketplace
# App Store Connect CLI (iOS devs)
brew install rudrankriyam/tap/asc
npx skills add rudrankriyam/app-store-connect-cli-skills
# Remotion (programmatic video)
npx skills add remotion-dev/skills
# Marketing Skills (32 skills for growth)
npx skills add coreyhaines31/marketingskillsA Note on Security
This matters, so I’m saying it plainly: a recent analysis found that over a third of community skills have security vulnerabilities. The eight skills in this list come from three sources I trust: Anthropic (official), obra/superpowers (99K+ stars, Anthropic marketplace verified), and well-known developers with established, auditable track records.
Before installing anything from an unknown source: read the SKILL.md. Check the allowed-tools field. If a skill needs Bash access and you don’t understand why, skip it. Skills are just markdown files with instructions - that simplicity is their power and their risk. Treat them like dependencies.
The tools are here. The ecosystem is mature enough to be useful and young enough that most developers haven’t caught up yet. That gap is your advantage.
Install the ones that match your workflow. Skip the rest. And if you haven’t read the architecture behind all of this - why Skills exist, how AGENTS.md shapes your agent’s behavior, and how Hooks enforce the rules your agent can’t ignore - the full series is here:
Ship faster. Ship better. Stop doing everything by hand.
Follow me for more practical AI coding workflows. I build apps, write about the tools I use, and share what actually works.