Show HN: VibeRaven – Production workflows for AI coding agents

VibeRaven, a new open-source tool for AI coding agents, provides production workflows that enforce evidence-based task completion, separating code fixes from provider dashboard work and requiring approval before risky changes. The tool includes a local Studio cockpit, plugin-style metadata, and skills for tasks like RLS design, webhook handling, and launch readiness.

Ship AI-built apps with evidence, not vibes. Open the full-quality MP4 demo VibeRaven is the VibeRaven Production Skills plus a local Studio cockpit for AI-built apps. Use it while building, fixing, reviewing releases, or preparing a launch when an agent needs provider context, release context, approval-aware chat, safe repo changes, and evidence before it claims the work is ready. npx -y viberaven The public repo is the agent discovery and installation surface: plugin-style metadata, portable slash commands, production skills, MCP notes, and AI-readable docs. AI coding agents are good at patching code and bad at knowing what they cannot prove. VibeRaven Production Skills are not passive reports; they are task workflows an agent can use while implementing, reviewing, debugging, or preparing launch-sensitive work. VibeRaven gives agents a contract: - say what evidence was found; - say what evidence is missing; - separate repo-code fixes from provider dashboard work; - use MCP or connected provider context when available; - keep release and version context visible; - ask for approval before risky local changes. That matters during feature work, auth fixes, billing changes, deploy debugging, release comparison, post-launch drift, and production readiness. Open the local cockpit from any project: npx -y viberaven Studio gives you agentic chat, draggable providers, draggable versions/releases, provider MCP context, terminal output, diff context, access-mode control, and CLI-agent connection checks. Installed is not the same as connected; test the agent connection before real chat control. Install bounded VibeRaven guidance for Codex, Claude Code, Cursor, Copilot, Gemini, and related agents: npx -y viberaven init --agents all Preview without writing files: npx -y viberaven init --agents all --dry-run The installer writes rules into supported agent instruction files, including AGENTS.md , CLAUDE.md , GEMINI.md , Cursor rules, Copilot instructions, and .viberaven context files. VibeRaven ships as a portable skill/plugin-style pack: - Codex: .codex-plugin/plugin.json - Claude Code: .claude-plugin/plugin.json - Gemini CLI: gemini-extension.json - Generic plugin hosts: plugin.yaml - Slash-command prompts: commands/ Command entry points: /viberaven-help : show the pack and output contract. /viberaven-work : use Production Skills on a task. /viberaven-launch : run launch readiness before saying ready. /viberaven-provider-actions : separate repo-code fixes from dashboard work. See docs/agent-portability.md /ohad6k/VibeRaven/blob/main/docs/agent-portability.md . Browse the skill library in docs/production-skills.md /ohad6k/VibeRaven/blob/main/docs/production-skills.md . | Skill | What It Helps Agents Do | |---|---| supabase-rls | Design or repair RLS work while separating repo SQL from live dashboard state. | stripe-webhooks | Implement safer webhook handling, idempotency, mode separation, and provider follow-up. | vercel-env-sync | Fix env assumptions across local, preview, and production without guessing dashboard values. | clerk-callbacks | Debug auth callback, redirect, preview URL, and route-protection drift. | sentry-signal | Wire or review monitoring so installed SDKs are not mistaken for working signal. | release-review | Review diffs and guide release-risk fixes for provider, auth, billing, env, data, and monitoring changes. | provider-actions | Split what the agent can edit in repo from what a human or MCP/provider tool must handle. | launch-readiness | Pull together implementation status, verification output, provider gaps, and next actions. | evidence-first | Keep claims honest when repo code, MCP context, or provider evidence is incomplete. | npx -y skills add ohad6k/VibeRaven --skill viberaven VibeRaven can be exposed to MCP-aware agents: { "viberaven": { "command": "npx", "args": "-y", "viberaven", "--mcp" } } Use MCP output as provider and readiness context for Studio-aware agents. The repo also keeps MCP registry metadata in docs/mcp-registry-submission.md /ohad6k/VibeRaven/blob/main/docs/mcp-registry-submission.md so maintainers can verify the public MCP discovery surface alongside npm and skills.sh metadata. examples/nextjs-supabase-vercel-production-ready-template /ohad6k/VibeRaven/blob/main/examples/nextjs-supabase-vercel-production-ready-template includes agent rules and viberaven: scripts for Next.js + Supabase + Vercel. llms-full.txt https://viberaven.dev/llms-full.txt llms.txt /ohad6k/VibeRaven/blob/main/llms.txt skills.json https://viberaven.dev/skills.json skills.sh.json /ohad6k/VibeRaven/blob/main/skills.sh.json Production Protocol guide https://viberaven.dev/viberaven-production-protocol-ai-built-apps.md Example evidence artifacts /ohad6k/VibeRaven/blob/main/examples/proof Older VibeRaven docs and compatibility tools may mention agent-mode scans, task lists, gate results, PRP resources, or scan-derived action manifests. Treat that language as compatibility context, not the main public product surface. The current default is the Studio: npx -y viberaven Normal git push is not gated. VibeRaven language about readiness is about production-change confidence, release review, provider-aware evidence, and agent boundaries, not blocking ordinary repository work. - Website: viberaven.dev https://viberaven.dev - npm: viberaven https://www.npmjs.com/package/viberaven - Issues: ohad6k/VibeRaven/issues https://github.com/ohad6k/VibeRaven/issues - Public discovery repo: ohad6k/VibeRaven https://github.com/ohad6k/VibeRaven