{"id":"2078137155424285163","url":"https://x.com/aiedge_/status/2078137155424285163","text":"","author":{"name":"AI Edge","username":"aiedge_","avatarUrl":"https://pbs.twimg.com/profile_images/1910271186442846208/vrYKSYRn_200x200.jpg"},"createdAt":"Fri Jul 17 15:18:02 +0000 2026","engagement":{"replies":5,"retweets":27,"likes":228,"views":134340},"article":{"title":"10 AI Skills So Powerful They Feel Illegal (install now) ","previewText":"The complete playbook for installing, deploying, and refining the 10 most powerful AI skills on the market.\nThese skills work in Hermes, Codex, Claude Code, and any other AI harness you could think","coverImageUrl":"https://pbs.twimg.com/media/HMkk9uca4AA_kwP.jpg","content":"## The complete playbook for installing, deploying, and refining the 10 most powerful AI skills on the market.\n\nThese skills work in Hermes, Codex, Claude Code, and any other AI harness you could think of.\n\nOver the past year, I've shared dozens of AI skills on this page and I've personally spent hours researching, testing, and running 50+ agentic skills.\n\nThis article is my list of the top 10 skills that have delivered genuine value to my AI workflows. \n\nNo slop, just pure alpha, and honestly, most of these skills probably should have been paywalled. \n\nBe sure to bookmark this now so you don't lose this list! \n\nShoutout to all the creators and contributors for these skills\n\nContents \n\nl: How To Install Each Skill \n\nll: AI Skills #10-#1\n\nlll: How to Refine These Skills\n\n## How To Install Each Skill\n\nBefore I reveal the top 10 skills list, there are two primary methods of installation I recommend:\n\n1. Terminal Command\n\nEach skill includes a terminal command you can paste into your AI harness to install it instantly.\n\nTo verify installation for each, type /skill (or a similar command) depending on your AI harness. \n\nIn the skills list below, I'll include the exact command for each.\n\n2.     Prompting\n\nThis is my personal preferred method, and the fastest way to get up and running if you're not deep in the terminal.\n\nWith each skill, you can simply tell your AI agent to install it for you by including the GitHub link to the respective skill. \n\nExample Prompt\n\nThe agent handles the entire end-to-end installation process (super simple).\n\nFor each skill, I've attached the direct GitHub link which you can send to your agent for easy Instillation. \n\nWith that out of the way, let's dive into the 10 AI skills you should be running (updated July 2026): \n\n## #10: /loopy \n\n![https://github.com/Forward-Future/loopy](https://pbs.twimg.com/media/HNMoo59aMAEDKk2.jpg)\n\nMost prompts ask an agent to do something once. Loopy gives your agent the ability to learn from the result and take the next useful step, turning a one-shot instruction into a repeatable, self-improving workflow.\n\nIt's two things in one: a public library of ready-made agent loops you can browse and copy, and an installable skill that gives your agent guided access to that library. \n\nWhat it does:\n\n- Discover: Scans your codebase or coding history for repeated work and turns the strongest candidates into loops\n\n- Find: Recommends published loops that fit your task\n\n- Loop Doctor: Audits existing loops \n\n- Adapt: Tailors any published loop to your personal workflows \n\n(and much more)\n\nInstall:\n\nClaude Code:\n\nCodex:\n\nExample prompts:\n\n/loopy Find a loop for [what you want to do]\n/loopy Interview me and help me craft a loop for [goal]\n/loopy Audit this loop and repair only material problems: [paste loop]\n\n## #9: Claude-Video \n\n![https://github.com/bradautomates/claude-video](https://pbs.twimg.com/media/HNMoo6BbQAAMs2F.jpg)\n\nEven with this skill in my #9 spot, it's one of my top picks, and it only recently came on my radar. \n\nThis skill enables Claude Code to \"watch\" videos.\n\nHow It Works \n\n→ Downloads from YouTube, Instagram, X, Vimeo, TikTok, Loom, and 1,800+ other sites via yt-dlp\n\n→ Extracts frames intelligently based on video length:\n\n→ Pulls a timestamped transcript from native captions for free\n\n→ Handles blurry, flipped, or noisy frames using bash and crop tools automatically\n\n→ Cleans up temp files when you're done\n\nInstall\n\nClaude Code (manual):\n\nCodex:\n\n## #8: /last30days\n\n![https://github.com/mvanhorn/last30days-skill](https://pbs.twimg.com/media/HNMoo6BacAA06R3.jpg)\n\nThis skill is for all marketers, social media managers, and even those trying to grow their own social media presence.\n\nThis AI skill searches trending social platforms (Reddit, X, YouTube, Hacker News, Polymarket, GitHub, TikTok, Instagram) and the broader web simultaneously, ranks everything by upvotes, likes, and real money, and synthesizes them into a single grounded brief. \n\nWhat makes it different from web search:\n\nTL;DR: Find out what's trending on the web in seconds \n\nInstall:\n\nAll harnesses (Codex, Cursor, Gemini CLI, and 50+ others):\n\nUse case examples \n\n- /last30days What is the current sentiment on physical AI and robotics stocks?\n\n- /last30days what are people saying about Claude Fable 5 right now?\n\n- /last30days what content formats are performing best on X this month?\n\n## #7: Kill AI Slop\n\n![https://github.com/hardikpandya/stop-slop](https://pbs.twimg.com/media/HNMoo58bYAAwRAb.png)\n\nThis skill kills all signs of AI slop in text.\n\nThe 8 categories it targets:\n\n- Banned phrases \n\n- Structural clichés \n\n- Passive voice \n\n- Wh- sentence starters \n\n- Em dashes \n\n- Vague declaratives \n\n- Narrator-from-a-distance voice \n\n- Rhythm patterns \n\nAll harnesses:\n\nPrompting this skill:\n\n\"Apply stop-slop rules to this text: [paste text].\"\n\n## #6: GOG (Google Workspace CLI)\n\n![https://clawhub.ai/steipete/skills/gog](https://pbs.twimg.com/media/HNMoo5-aUAAiQ-u.jpg)\n\nA skill that installs Google Workspace CLI in OpenClaw.\n\nWhat it covers\n\n- Gmail \n\n- Calendar \n\n- Drive \n\n- Sheets \n\n- Docs \n\n- Contacts\n\nOpenClaw Command\n\n## #5: Unslop UI\n\n![https://github.com/JCarterJohnson/vibecoded-design-tells](https://pbs.twimg.com/media/HNMoo6MbMAA51lA.jpg)\n\nThis skill is trained on thousands of \"AI slop\" datapoints.\n\nComments from real Redditors, social media comments & more.\n\n![data this skill is trained on](https://pbs.twimg.com/media/HNMoo6HagAA4qZ-.jpg)\n\nIt removes slop by targeting the typical AI slop patterns - purple/colorful icons, blocky logos, etc.\n\nIt has two modes \n\nBuild mode: You write what the design should be before any code gets generated.\n\nAudit mode: Runs the standalone scanner against an existing codebase. \n\nClaude Code Command \n\n## #4: Shannon Pentester\n\n![https://github.com/KeygraphHQ/shannon](https://pbs.twimg.com/media/HNMoo6WbsAAc5SU.jpg)\n\nEvery single vibe-coder should install this.\n\nShannon is a fully autonomous, white-box AI pentester for web apps and APIs. It actively probes your codebases for possible security vulnerabilities, attack vectors and security threats.\n\nInstallation is extremely easy \n\n## #3: Security Unbroker \n\n![https://github.com/NousResearch/hermes-agent/tree/main/optional-skills/security/unbroker](https://pbs.twimg.com/media/HNMoo6ZbkAAwRHT.jpg)\n\nUnbroker is a Hermes-native skill that automates the entire process of removing your personal identifiable information from the web.\n\nIt scans the broker registry, identifies where your data appears, submits opt-out requests, verifies the removals, schedules re-checks, and reports back. \n\nThe parts that require a human, hard CAPTCHAs, government ID verification, or phone calls, get escalated to a human task queue. Everything else runs hands-off.\n\nWhat it does:\n\n- Scans data brokers and people-search sites in parallel\n\n- Submits opt-out requests autonomously\n\n- Prioritizes cluster-parent deletions\n\n- Schedules re-checks automatically \n\n- Produces a full removal ledger and audit report\n\n- Manages multiple people independently (so you can run it for family members or clients, each isolated)\n\nHermes install\n\n## #2: /improve \n\n![https://github.com/shadcn/improve](https://pbs.twimg.com/media/HNMoo6faEAA0itI.jpg)\n\n/improve reads your entire codebase, identifies the highest-value improvements, and writes plans to improve your projects.\n\nThe four commands:\n\n- /improve: full codebase audit, writes all plans\n\n- /improve execute <plan>: dispatches a cheaper model to implement\n\n- /improve branch: scoped audit \n\n- /improve reconcile: verifies what landed\n\nInstall:\n\nClaude Code:\n\nAll harnesses:\n\n## #1: Taste Skill\n\n![https://www.tasteskill.dev/](https://pbs.twimg.com/media/HNMoo6yasAA8DUS.jpg)\n\nThis \"Taste\" skill is an anti-slop AI frontend framework.  \n\nFeed it to your Hermes agent, Claude Code, or Codex to get rid of generic AI slop. \n\nWorks for: Design skills, image generation, web/mobile app UI/UX & more.\n\nIt comes with seven variants:\n\n- design-taste-frontend: the core skill, covers everything\n\n- redesign-audit: upgrading existing projects\n\n- soft-ui: glass effects\n\n- minimalist: Notion and Linear-style interfaces\n\n- brutalist: Swiss typography, raw structure, sharper contrast\n\n- no-lazy: blocks placeholder comments, skipped sections, half-finished outputs\n\n- image-to-code: generates reference frames first, analyzes them, then implements\n\nInstall for all harnesses:\n\n## How To Refine These Skills\n\nInstalling a skill is step one. \n\nThe real value comes from refining it over time until it fits exactly how you work. Here's how to do that (general frameworks):\n\nIteration loops\n\nEvery skill gets better through the same cycle: \n\nRun it → give feedback → update the file → run it again\n\nAfter any skill run that produces output you're not happy with, you should open the SKILL.md file and add a rule. \n\nExample\n\n\"Hermes, I need you to update my taste skill with no blue color palettes - it looks like slop.\" \n\nThere are two primary ways that I like to refine my skills:\n\n1. Use /skill creator (Claude Code)\n\nThe fastest way to improve an existing skill. Open a Claude Code session after a run you want to improve and type:\n\n/skill creator\n\nThen, give it the skill file and your feedback.\n\n2. Run a dedicated refinement loop\n\nFor skills you use constantly, set up a loop that improves the skill automatically based on your feedback:\n\n/loop after every run of [skill name], ask me for feedback, update the SKILL.md with any new rules I give, and keep a changelog of every change made to the file.\n\nNow every time you use the skill and flag something, the improvement is written in permanently instead of getting lost.\n\nNote: It's also a good strategy to edit your skill files manually or attach them in a tool like Cowork to easily access them & update them at anytime.\n\n## Closing Thoughts \n\nI hope you found my list of the top 10 AI skills valuable.\n\nIf you did, be sure to follow me here @aiedge_ - every single week I post AI articles to help you get the edge in AI.\n\nIf you enjoy written content about AI, feel free to subscribe to my free newsletter. We send you weekly breakdowns of the latest AI news, practical guides, tips & more:\n\nhttps://newsletter.aiedgehq.co/\n\n![https://newsletter.aiedgehq.co/](https://pbs.twimg.com/media/HNMoo7LagAAtvHD.jpg)\n\n100% free, no spam ever & unsub anytime!\n\nThanks for reading💙"},"adhxContext":{"savedByCount":1,"publicTags":[],"previewUrl":"https://adhx.com/aiedge_/status/2078137155424285163"}}