VibeVault
Real-time API key leak prevention for Vibe Coding environments. Detects hardcoded secrets the moment you type — and fixes them in one click. Why VibeVault?The rise of AI coding assistants — ChatGPT, Claude, Cursor, Copilot — has dramatically lowered the barrier to shipping software. A developer with minimal experience can now deploy a live web service in hours. But speed without guardrails creates a structural security gap:
Public API key leak trackers aggregate hundreds of new exposures from public repositories every day — the vast majority originating from AI-assisted projects. This is no longer an edge case. VibeVault intercepts this problem at the IDE level. Before the code hits Git, VibeVault detects exposed secrets and guides you to fix them automatically. Features1. Real-time Detection (9 Service Patterns)Scanning triggers within 500ms of typing, covering the most common services used in AI-generated code.
2. One-click Auto-fix (QuickFix)Press What happens automatically:
3. Smart Variable Name InferenceVibeVault analyzes the surrounding code to suggest a meaningful variable name.
4. Security Status BarLive security indicator in the VS Code status bar (bottom-right):
Click to trigger a full workspace scan. 5. Workspace-wide Scan
6. Startup
|
| Setting | Type | Default | Description |
|---|---|---|---|
vibevault.enable |
boolean | true |
Enable or disable secret scanning |
vibevault.confirmVariableName |
boolean | true |
Show input box to confirm variable name before writing |
Design Principles
No network calls. All detection runs locally via regex. Your code never leaves your machine. Works offline.
Minimal false positives. Lines that already use process.env.KEY, os.getenv(...), ENV[...] etc. are skipped. Safe references on other lines do not suppress detection on the current line.
Non-destructive edits. All changes go through VS Code's WorkspaceEdit API — fully undoable with a single Ctrl+Z, including auto-inserted imports.
Roadmap
- Git pre-commit hook integration — block commits containing raw secrets
- Auto-generate
.env.examplefor team onboarding - Entropy-based detection — catch high-entropy strings beyond known patterns
- CI/CD integration — apply the same rules in GitHub Actions pipelines
Release Notes
0.1.2
- Fix: GitHub PAT pattern relaxed from
{36}to{35,}to match all token lengths
0.1.1
- Fix: Keys on lines below a
os.getenv()call were incorrectly suppressed —isAlreadySafenow scoped to current line only - Fix: Anthropic keys were being double-flagged by the OpenAI pattern — added negative lookahead
(?!ant-) - Fix: Variable name mismatch between
.enventry and code replacement when a naming conflict occurred - Fix: Duplicate detection in
.envchanged from substring match to exact line match - Feature: Python
import osauto-inserted as part of the same undoable action - Feature: Go
import "os"advisory message - Improvement:
inferEnvVarNamecorrectly handles keyword arguments (OpenAI(api_key="...")) - Improvement: 500ms debounce reduces CPU usage during active typing
0.1.0
- Expanded from 1 to 9 detection patterns (Anthropic, AWS, Google, GitHub, Stripe, HuggingFace, Generic)
- Fixed quote-inclusive code replacement
- Added status bar security indicator
- Added workspace scan command
- Added smart variable name inference
- Added startup
.gitignorevalidation - Added language-specific env var reference syntax for 7 languages
0.0.1
- Initial release: OpenAI
sk-pattern detection and.envmigration