Graphify, Ponytail, Playwright CLI: what the rising agent skills actually install

Three agent-skill names are climbing in search right now: Graphify, Ponytail and Playwright CLI. Each is more than a single SKILL.md. Graphify adds a hook or always-on rule, Ponytail is a plugin with lifecycle hooks, and Playwright CLI is a global npm tool that copies a skill folder into your project. Know which of those you are accepting before you paste the install line.

The demand signal

We checked Google Trends on 5 October 2026 (about 09:00 Asia/Riyadh), worldwide, past 30 days:

The same rising lists also contained obviously unrelated "Breakout" entries, so read this as direction, not volume. Trends values are relative; they are not search counts.

At a glance

Stars show attention, not safety.

Graphify: check the package name and the hook

• The PyPI package is `graphifyy` with two y's. The README says other graphify* packages on PyPI are not affiliated. The command you run afterwards is still graphify.
• On Claude Code it installs a PreToolUse hook that fires before search-style tool calls and nudges the agent to read the graph first. A --strict project mode goes further and blocks the first raw file read of a session.
• Code is parsed locally, according to the README, with tree-sitter and no model call. Docs, PDFs, images and video get a semantic pass through your assistant's model or an API key you configure, so those files leave your machine if you include them.
• Removal is one command: graphify uninstall, with --purge to also delete graphify-out/.

Ponytail: install only from the real repository

• The README says to install only from DietrichGebert/ponytail on GitHub or @dietrichgebert/ponytail on npm. It never ships `.exe` or `.dll` files; the author says a copy that does is not theirs.
• It is always on once installed, with lite, full (default) and ultra levels, plus commands such as /ponytail-review for a diff and /ponytail-audit for the whole repo.
• The headline savings in its README come from the author's own benchmark: 12 tasks, Haiku 4.5, n = 4. Treat them as a vendor claim to test on your own code.

Playwright CLI: a tool plus a skill, from Microsoft

• The skill documents the playwright-cli command surface: snapshots and element refs, sessions, request mocking, storage state, tracing and test generation.
• Project vs global: playwright-cli install --skills writes into the current workspace; -g writes into ~/.claude/skills or ~/.agents/skills for every project. Our project vs user scope post covers that choice.
• Node version: the repository README says Node.js 18 or newer; the playwright.dev installation page says 20 or newer. Use 20+ to satisfy both.
• Storage-state features save cookies and localStorage. Keep those files out of git.

Before installing any trending skill, answer three questions: *Is this the official source? What runs automatically — hooks, always-on rules, global files? How do I remove it?*

Which one fits which problem

• "My agent wastes turns grepping a big repo" → Graphify, scoped to one project first.
• "My agent over-engineers simple tasks" → Ponytail, ideally at lite before full.
• "My agent needs to test a web page" → the Playwright CLI skill, installed per project.

Read the install-hygiene checklist before adding any of them.