vibescore 0.1.0

Grade your vibe-coded project. One command, instant letter grade across security, quality, dependencies, and testing.

🎵 vibe-check

Grade your vibe-coded project. One command. Instant letter grade.

"Vibe coding" is the new reality — you prompt, AI writes, you ship.
But is your vibe-coded project actually good?
Find out in 10 seconds.

$ vibescore .

🎵 Vibe Check  v0.1.0
══════════════════════════════════════════════

  Project:   tokonomics
  Files:     40 (32 Python, 8 other)
  Lines:     4,658
  Scanned in 0.12s

┌──────────────────┬────────┬───────┐
│ Category         │ Score  │ Grade │
├──────────────────┼────────┼───────┤
│ Code Quality     │   52.0 │ F     │
│ Security         │  100.0 │ A+    │
│ Dependencies     │   98.0 │ A+    │
│ Testing          │  100.0 │ A+    │
├──────────────────┼────────┼───────┤
│ Overall          │   87.6 │ B+    │
└──────────────────┴────────┴───────┘

🟡 Warnings (11)
  VC201  Function 'export_svg_chart' too long (102 lines)
  VC202  Function '_build_cli' high complexity (30)
  VC203  Function 'export_svg_chart' has 6 parameters (>5)
  ...

💡 Tips
  • Reduce function complexity and add type annotations

Install

pip install vibescore

That's it. Zero dependencies. Works with Python 3.9+.

Usage

# Grade the current directory
vibescore .

# Grade a specific project
vibescore /path/to/project

# JSON output (for CI pipelines)
vibescore . --format json

# Fail CI if score is below threshold
vibescore . --min-score 70

As a Python library

from vibescore import scan

report = scan(".")
print(f"Grade: {report.overall_grade} ({report.overall_score:.0f}/100)")

for category in report.categories:
    print(f"  {category.name}: {category.grade}")

What It Checks

Category	Checks	Codes
Code Quality	Function length, cyclomatic complexity, parameter count, type annotations, nesting depth, star imports, docstrings, mutable defaults	VC201–VC209
Security	Hardcoded secrets, AWS keys, SQL injection, shell injection, unsafe deserialization, eval/exec, debug mode, private keys	VC301–VC309
Dependencies	Version pinning, lock files, deprecated setup.py, wildcard pins	VC401–VC405
Testing	Test file presence, test count, CI configuration, conftest.py, test-to-code ratio	VC501–VC506

Grading Scale

Grade	Score	Grade	Score
A+	97–100	C+	77–79
A	93–96	C	73–76
A-	90–92	C-	70–72
B+	87–89	D+	67–69
B	83–86	D	63–66
B-	80–82	D-	60–62
		F	0–59

CI Integration

GitHub Actions

- name: Vibe Check
  run: |
    pip install vibescore
    vibescore . --min-score 70

Pre-commit (manual)

# In your Makefile or CI script
vibescore . --min-score 70 --format json > vibe-report.json

How Scoring Works

Each category is scored 0–100 independently. The overall score is a weighted average:

Category	Weight
Security	30%
Code Quality	25%
Testing	25%
Dependencies	20%

Security is weighted highest because a security bug in vibe-coded projects can be catastrophic.

Why vibe-check?

Vibe coding means AI writes most of your code. That's fast, but it introduces risks:

• AI hallucinates long functions that are hard to debug
• AI skips security basics like input validation and secret management
• AI often omits tests or writes superficial ones
• AI uses loose dependency pins that break on updates

vibescore catches these patterns in seconds, so you can ship fast and ship safe.

FAQ

Q: Does this only work with Python?
A: Currently Python-focused for code quality and testing analysis. Security and dependency checks work with any project type. More languages coming soon.

Q: Does it phone home or require an API key?
A: No. Zero network requests. Zero dependencies. Runs entirely offline.

Q: How is this different from pylint/ruff/flake8?
A: Those are line-level linters. vibescore gives you a project-level grade across security, quality, testing, and dependencies — a holistic view of your vibe-coded project's health. Use both.

See Also

Tools in the same ecosystem:

• tokonomics — LLM token cost management
• injectionguard — Prompt injection detection
• vibesafe — AI code safety scanner
• castwright — Synthetic training data generator
• infermark — LLM inference benchmarking

License

Apache-2.0