Invocable Dispatch
A recurring, well-defined procedure only needs to be figured out once. Once it's written down — what it does, what it expects, what it produces — it gets a short, memorable, ALL-CAPS name, and typing that name is enough to run the whole documented procedure exactly, without re-explaining it or improvising a slightly different version each time.
The catalog is checked first — the risk this guards against is quietly reinventing, slightly differently, a procedure that already exists and is already documented.
Why a catalog instead of memory
Treating "I don't remember a procedure with this name" as proof that none exists is exactly the mistake this protocol is built to prevent. The catalog is the authority, not recall — an unfamiliar or newly added name gets a full catalog read before being dismissed as "not a real command." Only after that broader check comes back empty does a token get treated as an ordinary instruction in plain context.
What a match means
A matched invocable's specification is read in full and followed exactly — no summarizing the steps, no skipping ones that seem unnecessary this time, no substituting a similar-looking procedure instead. The specification is the source of truth for that procedure; treating it as a rough suggestion defeats the entire point of having named it in the first place.
Where new ones come from
When a genuinely new, repeatable procedure is needed and nothing in the catalog already covers it, a new invocable gets written up the same way — documented, named, registered — so the next time it's needed, it's a catalog lookup instead of a re-derivation.