Built two skills: /project-briefing and /klant-intake-dossier
So for Day 3 I went a bit further than one skill. Instead of writing a skill by hand, I used the skill-builder skill to interview me first (goal, trigger, steps, inputs/outputs, guardrails) and then build it. Building a skill with a skill — that clicked for me.
The first one, /project-briefing, summarizes any project folder into a 2-minute "get back in" briefing (goal, status, open points, next step). A sub-agent does the reading so my main chat stays clean.
The second one is the one I'm actually excited about: /klant-intake-dossier. It combines a skill with web scraping (Firecrawl). I give it a company name or URL, it scrapes the public website + socials, and returns a client intake dossier — including which of my services fits each opportunity. As a Business Support Partner that means I walk into a client conversation properly prepared, every time.
The biggest lesson wasn't the building, it was the iterating: run it, watch where it's slow or wrong, give feedback, run again. And understanding where a skill lives — global (works everywhere) vs project-level (only in that project) vs just stored in a library. Both skills are now live globally.
Process first, then AI. 😉