A Skill in Claude Code Is Convenient. Right-Click Without Claude Is Even Better
According to joint research by Qatalog and Cornell University (2021, Workgeist Report), the average office worker switches between apps about 1,200 times a day. Gloria Mark from UC Irvine measured that returning to a task after each such switch takes an average of 23 minutes.
→ It pays to solve tasks within a single interface.
That’s why I wrap most repetitive tasks into skills and handle them from the terminal in my IDE.
Fitts’s Law (Paul Fitts, 1954): the closer the target is to the cursor, the faster you can reach it. A context menu appears right at the cursor. Distance = 0. Physically, it doesn’t get faster than that.
→ And this is a relatively new approach for me (or maybe a long-forgotten old one) — adding commands to the right-click context menu. It’s incredibly convenient.
Two commands that used to live in my skills recently made it into the context menu as well.
LLMs constantly generate files in Markdown. On Mac I have QLMarkdown — an open-source app that renders .md files beautifully and error-free right in QuickLook. But when I share files with others, they don’t have it, and .md is a pain to read. So before sending, I convert to PDF. And honestly, I spent a long time trying to get quality .md-to-PDF rendering working — my research didn’t turn up a single working solution, so I built my own, though not on the first try.
Before, I could only render via a skill in the LLM (meaning I had to provide a file link, specify the skill, and ask it to make a PDF). Now I just right-click in the folder → Quick Actions → MD to PDF, 2–3 seconds, and the file is ready. It feels like magic in everyday life.
I have automated pipelines for transcribing recurring calls (plus my own code, optimized for M-series chips, for local transcription on my MacBook). For one-off calls from various apps, there was no pipeline — I’d have to open the Whisper UI, select a file, pick a model. Now it’s just right-click → Transcribe Audio, and I can watch the progress percentage in the terminal. Incredibly convenient.
Transcription loads about 5 GB of RAM, and I have a 16 GB MacBook — so I kick it off right before stepping away from the computer for 10 minutes, otherwise the machine starts lagging a bit. Usually an hour-long call transcribes in about 6 minutes.
Copy the repo links, paste them into your LLM, and ask it to create a skill and add it to the context menu. I use this on Mac, but I also asked Codex to adapt it for Windows — haven’t tested it though, there might be bugs. Experience the magic every day! 🪄
P.S. Apple in macOS Tahoe 26 is doing the same thing — embedding AI directly into Shortcuts and the context menu. Apparently I’m not the only one who thinks this is useful 🙂