blog

v1.5 — Build the AI app

Every Claude Code user is, by definition, AI-app-curious — and until now opchain had zero owners for the part of a project where an LLM is in the loop. v1.5 fixes that. It adds four AI-native skills and wires them through the pipeline you already use, so “the AI part” gets the same idea → spec → build → ship → evaluate treatment as the rest of the app.

The four new skills

The ripples

The new skills don’t sit off to the side — they extend what’s already there:

Dogfooding

opchain evaluates itself. The new prompts/opchain-eval/ set is a routing goldset — given a dev request, does opchain pick the right skill? — published as the worked example for /oc-prompt eval. We catch our own trigger-copy drift the same way we’d ask you to catch yours.

The theme of v1.5 is simple: building with AI should be an engineering discipline, not a vibe. Every new skill makes the AI part of your app an evaluated artifact — measured, versioned, and safe to ship.

Browse the skill library or install opchain to start.

← All posts