From one operator to a shared registry — a 12-month roadmap

The claim

The next 12 months take the catalog from one operator’s working library to a shared registry the top decile of LLM-using engineers can install and contribute to. Three milestones, each with a measurable close.

The evidence

The skill-export pipeline is already in place: any ready prompt emits an Anthropic-skill-format folder via one command. That is the substrate for the shared registry; it is not net-new work. The remaining 12 months land three milestones in sequence: a hosted index (Q3, target 1,000 anonymous installs), multi-model routing across Claude and one other frontier model (Q4, target 2 named integrations), and a paid hosted-library tier (Q1+1, target 100 paying operators at $10 per month).

Engineering capacity assumed: one founding engineer plus one hire from the round. No team-mode work in this window — the buyer remains the individual operator.

The comparable

Raycast followed the same sequence: free single-user tool first, public extension store second, paid pro tier third. Each step closed before the next opened. EVA is shipping the same shape, with the prompt artefact in the slot Raycast filled with the launcher extension.

What we ask for

A round sized to fund the two non-operator milestones (hosted index, paid tier) and one engineering hire. The thesis card frames the destination; the market card sizes the population; this card is the path.

timeline
  title 12-month roadmap from operator to registry
  Q3 2026 : Hosted public index : 1000 installs target
  Q4 2026 : Multi-model routing : Claude + 1 frontier model
  Q1 2027 : Paid hosted-library tier : 100 paying operators