hoard

name	hoard
description	Save a newly-figured-out trick as a minimal working example in hoard/, so future sessions can reuse or recombine it. Use after solving something non-obvious for the first time.

Hoard things you know how to do

We only ever need to figure out a trick once. If it's documented with a working code example, any future session can consult it and apply it to any similarly-shaped problem — or recombine two hoard entries into something new.

Recording an entry

The argument to this skill is the trick that was just figured out.

1. Create hoard/<slug>/README.md with:
   • What it does — one or two sentences, phrased as the question it answers ("Can a web page run OCR in JavaScript alone?").
   • The working example — the minimal complete code, in the README or as runnable files alongside it. It must be running code that was actually executed, not a theoretical sketch — knowing something is possible is not the same as having seen it work.
   • How to run it — the exact command(s).
   • Gotchas — anything that cost time to discover.
2. Keep it minimal: strip the example down to just the trick itself.

Using the hoard

• Before building something new, search hoard/ for existing examples to adapt.
• Recombine: the strongest prompting pattern is feeding two or more working examples into one prompt — "example A shows X, example B shows Y, combine them to build Z".
• External examples work too: fetch raw sources with curl (WebFetch summarizes; you usually want the raw HTML/code), or clone reference repos to /tmp and search them there.