bundled-skills/vibers-code-review/SKILL.md
Human review workflow for AI-generated GitHub projects with spec-based feedback, security review, and follow-up PRs from the Vibers service.
npx skillsauth add FrancoStino/opencode-skills-antigravity vibers-code-reviewInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
You push code. We review it against your spec, fix issues, and send a PR.
Use this skill when:
Go to your repo → Settings → Collaborators → Add marsiandeployer
Create .github/workflows/vibers.yml:
name: Vibers Code Review
on:
push:
branches: [main]
jobs:
review:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 2
- uses: marsiandeployer/vibers-action@v1
with:
spec_url: 'https://docs.google.com/document/d/YOUR_SPEC_ID/edit'
telegram_contact: '@your_telegram'
| Parameter | What it does |
|-----------|-------------|
| spec_url | Link to your spec (Google Doc, Notion, etc.). Must be publicly accessible (or "anyone with the link can view"). Without access to spec, review is impossible. |
| review_scope | full (default), security, or spec-compliance |
| telegram_contact | Your Telegram — we'll message you when review is ready |
Add this block to your project's CLAUDE.md, .cursorrules, or AGENTS.md:
## Commit messages
Every commit MUST include a "How to test" section in the body:
- Live URL to open and verify the change
- Step-by-step what to click/check
- Test credentials if login is required
- Expected result for each step
Example:
feat: Add user registration form
How to test:
- Open https://myapp.vercel.app/register
- Fill in email/password, submit
- Check that confirmation email arrives
- Try submitting with invalid email — should show error
- Login: [email protected] / demo123
Without "How to test" the reviewer has to guess what to verify, and the review takes longer.
Done. Now every push triggers a notification. You'll get a PR with fixes, usually within 24 hours.
We check: spec compliance, security (OWASP top 10), AI hallucinations (fake APIs/imports), logic bugs, UI issues.
We don't check: code style (use ESLint/Prettier), performance benchmarks, full QA (use Playwright/Cypress).
marsiandeployer as a collaborator| Plan | Rate | Details | |------|------|---------| | Promo | $1/hour | Full review + PRs with fixes. We ask for honest feedback in return. | | Standard | $15/hour | Full review + security audit + priority turnaround. |
No subscriptions. No contracts. Pay per review.
Send feedback directly from your agent:
curl -X POST https://vibers.onout.org/feedback \
-H 'Content-Type: application/json' \
-d '{"message": "Your question or issue", "repo": "https://github.com/you/your-repo"}'
Both message and repo are required. Response: {"status": "accepted"}.
Contacts:
Do I need an API key? No. Add collaborator + action, that's it.
What languages? JS/TS, Python, React, Next.js, Django, Flask, and more. If it's on GitHub, we review it.
What if I disagree with a fix? Comment on the PR. We discuss and adjust.
Can I use this without GitHub? Yes — write to Telegram with your code and spec.
data-ai
Snapshot a site's SEO state and detect ranking, indexation, metadata, canonical, robots, schema, and on-page regressions over time.
development
Coordinate focused subagents on substantial work, keep their ownership non-overlapping, and integrate verified results. Use for large-scope Codex tasks; keep trivial work with the coordinator.
data-ai
Use when an owner asks to find a cofounder or project partner. Assess only that agent's own owner and rank only approved profiles other agents posted for their own owners.
devops
Install, configure, verify, repair, update, and uninstall Hyprland on Fedora Linux with GPU-aware detection (NVIDIA/AMD/Intel).