Notes

Three skill libraries for Claude Code, each tackling a different angle.

Ponytail

The laziest senior dev lens. Flags reinvented stdlib, unneeded dependencies, speculative abstractions, and dead flexibility, one line per finding. Shortcuts taken get marked in code with a comment naming the upgrade path. Claims 80-94% less code, 42-75% less cost, 3-6x faster.

shadcn /improve

shadcn's own skills repo. The /improve skill audits a codebase by spawning parallel read-only Explore subagents, then hands a plan to a cheaper model to execute. Other skills give Claude project-aware shadcn context: framework, aliases, installed components.

NVIDIA SkillSpector

A security scanner for skills before you install them. Catches vulnerable dependencies, suspicious scripts, prompt injection, tool poisoning, and mismatches between a skill's declared purpose and its actual behavior. NVIDIA found 26.1% of skills have vulnerabilities and 5.2% look malicious.

Links

github.com

github.com

github.com

AI web design platform with a library of 400+ pre-built animated templates: hero sections, SaaS, portfolios, agencies, ecommerce. Freemium with a Go Unlimited upgrade. Integrates with Bolt for code generation and export. Powered by Design Rocket.

Useful for grabbing animated hero section inspiration without starting from scratch. The Bolt integration is the nicer part: a picked template exports as code instead of locking you into a builder, so the iteration can continue in your own stack.

Links

motionsites.ai

Reader-sourced reading list from Lenny's newsletter, grouped by what you want to get better at.

Design

• Don't Make Me Think, by Steve Krug
• The Design of Everyday Things, by Don Norman
• Refactoring UI, by Adam Wathan and Steve Schoger

Taste and craft

• The War of Art, by Steven Pressfield
• The Work of Art, by Adam Moss
• Creativity, Inc., by Ed Catmull

Influence

• How to Win Friends and Influence People, by Dale Carnegie
• Influence, by Robert Cialdini
• Never Split the Difference, by Chris Voss

Starting a company

• The Lean Startup, by Eric Ries
• Crossing the Chasm, by Geoffrey Moore
• Fall in Love with the Problem, Not the Solution, by Uri Levine

Career

• Great at Work, by Morten T. Hansen
• 7 Rules of Power, by Jeffrey Pfeffer
• The Effective Executive, by Peter Drucker

Happiness

• The Subtle Art of Not Giving a F*ck, by Mark Manson
• A Guide to the Good Life, by William B. Irvine
• Stumbling on Happiness, by Daniel Gilbert

Source

Miso-TTS (announced by @AodenTeoMT at MisoLabs) is an 8B-parameter text-to-speech model focused on emotive, human-like delivery. The weights are open-sourced.

Why it's interesting

• 110 ms latency — compared to ~300–700 ms for most competitors. Fast enough to feel real-time.
• One-shot voice cloning from a 10-second sample.
• Emotive output — the demo emphasizes emotional expressiveness, not just intelligibility.
• Open weights on HuggingFace (MisoLabs/MisoTTS) — designed for local deployment.
• English only at launch.

Install locally

Repo: github.com/MisoLabsAI/MisoTTS

Requirements: Python 3.10, CUDA GPU with enough VRAM for an 8B model in bfloat16 (so ~16GB+).

# 1. Install uv (or skip and use pip)
curl -LsSf https://astral.sh/uv/install.sh | sh

# 2. Clone and set up
git clone https://github.com/MisoLabsAI/MisoTTS.git
cd MisoTTS
uv sync --python 3.10
source .venv/bin/activate

# 3. Run inference (downloads weights from HuggingFace on first run)
uv run python run_misotts.py

Output lands at full_conversation.wav. Sony's SilentCipher watermarking model is auto-downloaded as part of the pipeline.

Pip alternative:

python3.10 -m venv .venv
source .venv/bin/activate
pip install -e .
python run_misotts.py

Things to try

• Drop in a 10-second clip of my own voice and see how good the clone is.
• Test latency end-to-end on my GPU (claimed 110 ms is the model alone, not network).
• Compare side-by-side with ElevenLabs / OpenAI TTS on the same script.

faster.dev/skills — a catalogue of reusable AI prompt “skills” (slash-command style) for coding tasks: audits, tests, architecture, framework init. Worth borrowing patterns from for any prompt library.

I knew many things would get replaced by AI (my businesses too) but human authenticity can't be, so I slowly started to focus on that a bit more, writing more fun blog posts here about my travels and things I do with my tech stack

— Pieter Levels (@levelsio)

Vibe Designing from the Terminal

Found a Claude Code skill that lets you design mobile apps straight from the terminal. Add one skill and you get a full vibe design + code workflow — no switching to Figma, no context loss. Saw someone on X showcasing their mobile app design workflow with it.

Links

• Sleek Design

Notion Skills

Notion shipped custom skills — reusable AI commands you build from any page. Turn a page into a skill via Use with AI → Use as AI skill, then invoke it from the text selection menu or @mention it in agent chats. Since skills are just pages, the whole team shares and iterates on the same version. Business and Enterprise plans only.

Links

• Notion Skills release

Intent Engineering

Context engineering tells agents what to know. Intent engineering tells agents what to want. It's the practice of encoding organizational purpose into infrastructure that agents can act against. […]

Intent engineering is how you communicate goals and objectives to the agent that allow the agent to work autonomously for long periods of time in a direction consistent with company strategy.

— @NateBJones on YouTube

AI won't kill professionals

"AI won't kill professionals. It will make the ones with judgment and taste impossible to replace. The question isn't whether AI takes your job. The question is whether you're good enough that people choose to work with you when they don't have to."

— Eric Siu on X

Links

github.com

x.com

Claude Code makes slides

Turns out Claude Code is surprisingly good at creating slide decks through web-based presentation libraries. Point it at something like Reveal.js or Slidev and it handles layout, content structure, and styling without much hand-holding. Faster than doing it manually and the output is cleaner than expected.

Links

quiver.ai