Audit UI quality across 8 dimensions and catch AI sameness. Scores design system compliance, visual hierarchy, spacing, contrast, typography, responsive, interactions, and accessibility. Returns a scorecard with priority-ordered fixes. TRIGGER on "vibe check", "UI audit", "design review", "does this look right"…
Personal operator layer. /op:go routes any ask; /op:spec /op:plan /op:build run the scaled gate for features; /op:ship /op:sync /op:show /op:release run the daily ops. Every command ends with proof.
Use when any work ask arrives — classifies it (bug / small fix / feature), confirms the route, and hands off to debugging, direct build, or the spec gate.
Use when an op command needs a per-project operation (serve, build, test, deploy, sync, release, live URL) — reads, discovers, and heals .claude/op.json in the target repo.
Use when work should go live — commit, push, deploy per recipe, cache-busted live verify, smoke, screenshot. Blocks the "shipped" claim on the live check.
Use when a change needs visual proof — serve locally per recipe, screenshot the standard responsive breakpoints (mobile, tablet, desktop), short verdict.
Use when starting any conversation — the op operator rules. Route work through /op:go, quote a repo convention before code, prove every claim, print the git status line after tree changes.
Llama Steve analyzes your actual product to identify and ship game-changing features. Maximize your conversion, cut user friction, and focus only on what wins.
AGENTS.md instructions for artttj/llama-steve, covering agents.md — llama steve, what this is, architecture (read before changing anything), running it and the teardown (what the agent does).