RTL is not a mirror, sorry
A mirrored file is not an RTL design. What flips, what stays, and who checks it.
Writing my process down once, so AI follows it every time
A "skill" is a saved set of instructions the AI loads for a type of task. I write my own taste, my own rules and my own past mistakes into them, once, so I do not have to repeat myself on the next project.
A skill is a saved set of instructions the AI loads before it does a certain kind of task. Not a personality, not a vibe. A written process: when to use it, the steps in order, and the specific rules I want followed.
I did not write these to make the AI smarter in general. I wrote them because I kept explaining the same thing project after project, and I got tired of repeating myself.
My launch-video workflow is a fixed order: brief and rules first, then real reference videos, then a text-only beat sheet, then one style-test shot before touching the rest, then the full draft, then one batched round of review, then render. I wrote it after a real launch film where jumping straight to visuals cost more rounds than it should have. Now the AI will not skip to visuals before the beat sheet is approved.
Reading real user feedback for my own tools follows a fixed procedure too: search for only what has not been read yet, summarize it the same way every time, mark it read, log it, then bring it to me to discuss. It never re-reads processed mail and never improvises the format, on purpose. Consistency matters more than cleverness for something I read every week.
Mood and tone still get made by hand first. A skill only carries the system-side rules: naming, order, structure, the mistake I am not making twice. And a skill is only as good as my own process was when I wrote it. If my flow is not solid, writing it into a skill just repeats the same mistake faster. I still read the output before it goes anywhere near a client.
Email me, or follow on LinkedIn where new notes land first. No newsletter, no pop-up, promise.