Blog & Skills

Notes, opinions, and copy-paste-ready skills.

Demo montage showing reanimate recording, speccing, building, and comparing a web animation side by side

reanimate: reverse-engineer any animation and prove the copy moves the same

A Claude Code skill that records a live web animation frame by frame, generates a timing spec, builds a copy in your project's stack, then loops compare-and-fix until the motion is verified.

Nov 5, 2026 · 2 min read

skill

skill · curated

chrome-devtools-mcp: Give Your Coding Agent Real Eyes in Chrome

The official Chrome DevTools team ships an MCP server that gives your coding agent live browser access — network inspection, performance traces, screenshots, and Puppeteer automation without you touching DevTools.

Nov 2, 2026 · 2 min read

skill

skill · curated

cloudflare/security-audit-skill: the agent auditor that argues with itself

Cloudflare's open-source skill turns your coding agent into a six-phase security auditor — the verifier that checks each finding is never the agent that found it.

Oct 29, 2026 · 2 min read

TasteCode banner showing the taste-skill desktop interface for AI-directed web design

skill · curated

taste-skill: give your coding agent an actual art director

A Claude Code skill that forces your agent to pick a real visual direction, source public-domain photography, and browser-check its own layout — instead of defaulting to rounded cards on a white background.

Oct 26, 2026 · 2 min read

snyk agent-scan v0.6 report showing scored MCP server and skill risks

skill · curated

snyk/agent-scan: a security scanner for the MCP servers and skills you've already installed

Snyk's CLI scanner auto-discovers your agent configs, inventories every MCP server and skill, and scores each for prompt injection, credential leaks, and destructive capabilities.

Oct 22, 2026 · 2 min read

skill

skill · curated

excalidraw-diagram-skill: your agent finally draws diagrams that argue their point

A Claude Code skill that generates Excalidraw architecture diagrams from natural language — and runs a Playwright render check to catch its own layout mistakes before you see them.

Oct 19, 2026 · 2 min read

skill

skill · curated

task-observer: the skill that watches your sessions and improves your whole toolkit

Most skills tell your agent what to do. task-observer watches what it actually does — then proposes fixes to the skills that let it down, and files new skill candidates for the patterns you keep repeating.

Oct 15, 2026 · 2 min read

PM Skills marketplace overview showing all nine plugins at a glance

skill · curated

pm-skills: 69 PM frameworks your AI agent can actually run

A 26k-star Claude Code skill marketplace that bakes Teresa Torres and Marty Cagan methodologies into nine installable plugins — from Opportunity Solution Trees to sprint pre-mortems.

Oct 12, 2026 · 2 min read

skill

skill · curated

graphify: turn your codebase into a queryable knowledge graph

A /graphify skill that turns any codebase—code, docs, SQL schemas, and PDFs—into a queryable knowledge graph using tree-sitter AST parsing, with no vector store or cloud dependency.

Oct 8, 2026 · 2 min read

skill

skill · curated

claude-skills: One Library, Thirteen AI Coding Agents

388 production-ready skills that install across Claude Code, Cursor, Windsurf, Aider, and ten more tools — plus a built-in security auditor so you actually know what you're running.

Oct 5, 2026 · 2 min read

3D graph visualization of a codebase knowledge graph in the codebase-memory-mcp web UI

skill · curated

codebase-memory-mcp: Structural Code Intelligence Without the Token Burn

An MCP server that indexes your entire codebase into a persistent knowledge graph using tree-sitter, answering structural queries in under 1ms at 99% fewer tokens than file-by-file reading.

Oct 1, 2026 · 2 min read

skill

skill · curated

mattpocock/skills: Pocock's Discipline Kit for AI-Assisted Dev

TypeScript educator Matt Pocock built a 271k-star collection of agent skills that addresses the specific ways AI pair programming goes wrong — agents that wander, produce unrunnable code, and quietly rot your architecture.

Sep 28, 2026 · 2 min read