VibeRaven

VibeRaven Agent Skills

Skills that teach AI coding agents to use VibeRaven — the control layer for AI-built products. Everything runs locally: no login, no API key, no telemetry.

Install with the Agent Skills CLI:

npx -y skills@1.7.0 add ohad6k/VibeRaven --skill vercel-supabase-launch-check
npx -y skills@1.7.0 add ohad6k/VibeRaven --skill viberaven
npx -y skills@1.7.0 add ohad6k/VibeRaven --skill architecture-context
npx -y skills@1.7.0 add ohad6k/VibeRaven --skill architecture-plan
npx -y skills@1.7.0 add ohad6k/VibeRaven --skill what-broke
npx -y skills@1.7.0 add ohad6k/VibeRaven --skill production-context
npx -y skills@1.7.0 add ohad6k/VibeRaven --skill go-live

The skills

vercel-supabase-launch-check: the pre-launch entry point. Triggers on the questions people actually ask before shipping a Vercel + Supabase app (tables missing RLS, a leaked service role key, env vars that differ in production, a Stripe webhook failing after deploy), runs the free local check, and explains each finding without calling the app secure.

viberaven: the router. Teaches agents the local loop: run npx -y viberaven@1.6.7 check for an offline repository check, read .viberaven/ artifacts, apply fixes with npx -y viberaven@1.6.7 fix --gap <id>, and check again once per batch of fixes. It is advice, not a gate: the user decides when to ship. Points at the Studio (npx -y viberaven@1.6.7) when the user wants to see and control the product.

architecture-context — the question gate. For vague feature work, asks the missing low-level product questions before any edit, then hands the answers to architecture-plan.

architecture-plan — the plan gate. Turns answered product questions plus repo evidence into a file-based workstream architecture plan before implementation.

what-broke — version context. Stops agents from patching blind when a working app regressed. Builds release context from git tags, diffs, and changelogs, then connects the change to provider reality (database, storage, deployment, runtime).

production-context — production memory. Keeps a compact .viberaven/production-context.md of boundaries, incidents, migrations, and fragile paths so the next agent action starts from what is dangerous, not just what exists.

go-live — the launch path. Takes a local app from repo state to GitHub remote to Vercel deployment with explicit build, push, deployment, and live-URL proof.

Each skill can end with Next skill:; continue with that skill unless user input, auth, or provider proof is required.

The loop agents follow

npx -y viberaven@1.6.7 check          # offline checks, 🔴/🟡/⚪ verdict, CI exit codes
npx -y viberaven@1.6.7 fix            # list gaps with safe automatic recipes
npx -y viberaven@1.6.7 fix --gap <id> # apply one recipe
npx -y viberaven@1.6.7 --strict       # the verdict as an exit code for CI, if you want one

Agents read .viberaven/agent-tasklist.md, .viberaven/gate-result.json, and .viberaven/context-map.json, fix one repo-code gap, then re-run check. Scans are local and free.

For Vercel + Supabase launch checks:

npx -y viberaven@1.6.7 audit --vercel-supabase