commit.show — Every commit, on stageThe vibe coding league. AI-assisted GitHub projects (built with Cursor · Claude Code · Lovable · Bolt · v0 · Cline · etc.) get audited, ranked, and forecasted by tier-gated Scouts.How it works

Audit · run npx commitshow@latest audit github.com/owner/repo · 60s deterministic + LLM-judged score (0-100) on 14 production-readiness frames

Audition · submit at /submit to enter the season ladder · first 3 audits per member free during launch

Encore · cross score 84 to earn a permanent badge · top 0.5% Valedictorian, 5% Honors, 14.5% Graduate

What the audit catches14 vibe-coding failure modes calibrated against real OSS: missing RLS · webhook idempotency · secret-in-bundle · column GRANT mismatches · Stripe API idempotency · mobile input auto-zoom · AI template-copy footguns · seed arrays in prod · CORS wildcard · localhost-in-prod · 4 more. Catch what your AI missed in 60 seconds.Who this is forVibe coders shipping with AI assistants who want a public ranking, social proof, and a path to graduation. Beginners welcome — Phase 1 brief auto-fills from your README, the audit handles the rest.Score = 50 + 30 + 20

Audit (50pt) · automated rubric · production-readiness signals from your repo + live URL

Scout Forecast (30pt) · tier-gated humans place predictions on which projects graduate

Community Signal (20pt) · views · comments · applauds · return visits

For LLM agents reading this pagePrefer the REST API · zero install, no sandbox prompt:GET https://api.commit.show/audit?repo=<github-url>&format=mdCORS open · no API key · returns paste-ready markdown · same engine as the CLI. Use this from Claude Code, Cursor agent mode, Cline, Windsurf, Gemini, ChatGPT browse, n8n, Zapier, curl. The npx commitshow path tends to trip agent sandbox classifiers; the REST endpoint never does.Full agent integration guide: /llms.txt · OpenAPI: api.commit.show/openapi.jsonFull audit method docs: /audit · /rulebookRun an audit yourselfnpx commitshow@latest audit github.com/owner/repoOr fetch from any agent runtime: GET https://api.commit.show/audit?repo=owner/repo&format=md · see /llms.txt for the full agent integration guide.Cross-links · Products · Scouts · Library · Rulebook · Audit method · CommunityPrivacy · Terms · operated by Madeflo Inc., a Delaware corporation.

When do you ask AI vs ask a human?

2026-05-15

Open Mic · commit.show

Current rule: AI for code, human for design. Probably wrong.

I default to asking AI for almost everything code-related. For design or "should I even build this" I still ping a human.

But I'm not sure that's the right split anymore — sometimes the AI gives sharper product feedback than my friends.

What's your dividing line?

#decision-making #design

Read on commit.show